Manolis Kellis: Biology of Disease - Lex Fridman 知识图谱

Lex Fridman (02:28.420)

to completely rethink how we engineer computational systems.

Manolis Kellis (02:33.560)

If you enjoy this thing, subscribe on YouTube, review it with 5 stars on Apple Podcast, follow

Lex Fridman (02:38.520)

on Spotify, support on Patreon, or connect with me on Twitter at Lex Friedman.

Lex Fridman (02:44.880)

And now, here's my conversation with Manolis Callas.

Lex Fridman (02:49.760)

So your group at MIT is trying to understand the molecular basis of human disease.

Lex Fridman (02:54.880)

What are some of the biggest challenges in your view?

Lex Fridman (02:57.720)

Don't get me started.

Manolis Kellis (02:58.720)

I mean, understanding human disease is the most complex challenge in modern science.

Lex Fridman (03:06.240)

So because human disease is as complex as the human genome, it is as complex as the

Manolis Kellis (03:13.340)

human brain, and it is in many ways, even more complex because the more we understand

Manolis Kellis (03:20.680)

disease complexity, the more we start understanding genome complexity and epigenome complexity

Lex Fridman (03:27.000)

and brain circuitry complexity and immune system complexity and cancer complexity and

Lex Fridman (03:31.200)

so on and so forth.

Lex Fridman (03:32.280)

So traditionally, human disease was following basic biology.

Manolis Kellis (03:39.400)

You would basically understand basic biology in model organisms like, you know, mouse and

Manolis Kellis (03:44.360)

fly and yeast.

Manolis Kellis (03:46.040)

You would understand sort of mammalian biology and animal biology and eukaryotic biology

Manolis Kellis (03:53.400)

in sort of progressive layers of complexity, getting closer to human phylogenetically.

Lex Fridman (03:59.920)

And you would do perturbation experiments in those species to see if I knock out a gene,

Lex Fridman (04:06.640)

what happens?

Lex Fridman (04:07.960)

And based on the knocking out of these genes, you would basically then have a way to drive

Manolis Kellis (04:12.720)

human biology because you would sort of understand the functions of these genes.

Lex Fridman (04:16.880)

And then if you find that a human gene locus, something that you've mapped from human genetics

Manolis Kellis (04:23.660)

to that gene is related to a particular human disease, you'd say, aha, now I know the function

Lex Fridman (04:28.760)

of the gene from the model organisms.

Manolis Kellis (04:31.440)

I can now go and understand the function of that gene in human.

Lex Fridman (04:37.120)

But this is all changing.

Manolis Kellis (04:38.320)

This is dramatically changed.

Lex Fridman (04:39.440)

So that was the old way of doing basic biology.

Manolis Kellis (04:41.760)

You would start with the animal models, the eukaryotic models, the mammalian models, and

Lex Fridman (04:46.280)

then you would go to human.

Manolis Kellis (04:48.680)

Human genetics has been so transformed in the last decade or two that human genetics

Lex Fridman (04:55.320)

is now actually driving the basic biology.

Manolis Kellis (04:58.420)

There is more genetic mutation information in the human genome than there will ever be

Lex Fridman (05:04.300)

in any other species.

Lex Fridman (05:06.240)

What do you mean by mutation information?

Lex Fridman (05:08.440)

So perturbations is how you understand systems.

Lex Fridman (05:11.380)

So an engineer builds systems and then they know how they work from the inside out.

Lex Fridman (05:16.120)

A scientist studies systems through perturbations.

Lex Fridman (05:20.360)

You basically say, if I poke that balloon, what's going to happen?

Lex Fridman (05:23.040)

And I'm going to film it in super high resolution, understand, I don't know, aerodynamics or

Manolis Kellis (05:26.480)

fluid dynamics if it's filled with water, et cetera.

Lex Fridman (05:28.840)

So you can then make experimentation by perturbation and then the scientific process is sort of

Manolis Kellis (05:33.960)

building models that best fit the data, designing new experiments that best test your models

Lex Fridman (05:41.120)

and challenge your models and so on and so forth.

Manolis Kellis (05:43.320)

This is the same thing with science.

Manolis Kellis (05:44.800)

Basically if you're trying to understand biological science, you basically want to do perturbations

Manolis Kellis (05:49.440)

that then drive the models.

Lex Fridman (05:54.600)

So how do these perturbations allow you to understand disease?

Lex Fridman (05:58.320)

So if you know that a gene is related to disease, you don't want to just know that it's related

Lex Fridman (06:04.120)

to the disease.

Manolis Kellis (06:05.120)

You want to know what is the disease mechanism because you want to go and intervene.

Lex Fridman (06:09.960)

So the way that I like to describe it is that traditionally epidemiology, which is basically

Manolis Kellis (06:17.240)

the study of disease, you know, sort of the observational study of disease has been about

Lex Fridman (06:23.000)

correlating one thing with another thing.

Lex Fridman (06:25.760)

So if you have a lot of people with liver disease who are also alcoholics, you might

Manolis Kellis (06:29.880)

say, well, maybe the alcoholism is driving the liver disease or maybe those who have

Manolis Kellis (06:34.360)

liver disease self medicate with alcohol.

Lex Fridman (06:36.780)

So the connection could be either way.

Manolis Kellis (06:40.120)

With genetic epidemiology, it's about correlating changes in genome with phenotypic differences

Lex Fridman (06:47.640)

and then you know the direction of causality.

Lex Fridman (06:50.120)

So if you know that a particular gene is related to the disease, you can basically say, okay,

Lex Fridman (06:58.120)

perturbing that gene in mouse causes the mice to have X phenotype.

Lex Fridman (07:03.860)

So perturbing that gene in human causes the humans to have the disease.

Lex Fridman (07:08.240)

So I can now figure out what are the detailed molecular phenotypes in the human that are

Manolis Kellis (07:14.680)

related to that organismal phenotype in the disease.

Lex Fridman (07:18.820)

So it's all about understanding disease mechanism, understanding what are the pathways, what

Manolis Kellis (07:22.960)

are the tissues, what are the processes that are associated with the disease so that we

Lex Fridman (07:27.560)

know how to intervene.

Manolis Kellis (07:29.000)

You can then prescribe particular medications that also alter these processes.

Manolis Kellis (07:33.360)

You can prescribe lifestyle changes that also affect these processes and so on and so forth.

Manolis Kellis (07:37.920)

That's such a beautiful puzzle to try to solve.

Manolis Kellis (07:41.040)

Like what kind of perturbations eventually have this ripple effect that leads to disease

Manolis Kellis (07:45.480)

across the population.

Lex Fridman (07:46.480)

And then you study that for animals or mice first and then see how that might possibly

Manolis Kellis (07:51.880)

connect to humans.

Lex Fridman (07:54.680)

How hard is that puzzle of trying to figure out how little perturbations might lead to,

Lex Fridman (08:01.340)

in a stable way, to a disease?

Lex Fridman (08:04.200)

In animals, we make the puzzle simpler because we perturb one gene at a time.

Manolis Kellis (08:11.040)

That's the beauty of this, the power of animal models.

Lex Fridman (08:13.500)

You can basically decouple the perturbations.

Manolis Kellis (08:15.800)

You only do one perturbation and you only do strong perturbations at a time.

Manolis Kellis (08:21.240)

In human, the puzzle is incredibly complex because obviously you don't do human experimentation.

Manolis Kellis (08:28.600)

You wait for natural selection and natural genetic variation to basically do its own

Manolis Kellis (08:34.720)

experiments, which it has been doing for hundreds and thousands of years in the human population

Lex Fridman (08:40.560)

and for hundreds of thousands of years across the history leading to the human population.

Lex Fridman (08:49.320)

So you basically take this natural genetic variation that we all carry within us.

Manolis Kellis (08:54.440)

Every one of us carries 6 million perturbations.

Lex Fridman (08:58.280)

So I've done 6 million experiments on you, 6 million experiments on me, 6 million experiments

Manolis Kellis (09:02.920)

on every one of 7 billion people on the planet.

Lex Fridman (09:06.400)

What's the 6 million correspond to?

Manolis Kellis (09:08.600)

6 million unique genetic variants that are segregating in the human population.

Manolis Kellis (09:14.840)

Every one of us carries millions of polymorphic sites, poly, many, morph, forms.

Manolis Kellis (09:22.880)

Polymorphic means many forms, variants.

Manolis Kellis (09:25.080)

That basically means that every one of us has single nucleotide alterations that we

Manolis Kellis (09:29.560)

have inherited from mom and from dad that basically can be thought of as tiny little

Lex Fridman (09:34.680)

perturbations.

Manolis Kellis (09:36.320)

Most of them don't do anything, but some of them lead to all of the phenotypic differences

Lex Fridman (09:42.480)

that we see between us.

Manolis Kellis (09:43.920)

The reason why two twins are identical is because these variants completely determine

Lex Fridman (09:48.960)

the way that I'm going to look at exactly 93 years of age.

Lex Fridman (09:52.520)

How happy are you with this kind of data set?

Lex Fridman (09:54.720)

Is it large enough of the human population of Earth?

Lex Fridman (09:59.240)

Is that too big, too small?

Lex Fridman (10:01.680)

Yeah, so is it large enough is a power analysis question.

Manolis Kellis (10:07.360)

In every one of our grants, we do a power analysis based on what is the effect size

Lex Fridman (10:11.440)

that I would like to detect and what is the natural variation in the two forms.

Manolis Kellis (10:19.200)

Every time you do a perturbation, you're asking, I'm changing form A into form B. Form A has

Manolis Kellis (10:25.240)

some natural phenotypic variation around it and form B has some natural phenotypic variation

Manolis Kellis (10:30.160)

around it.

Manolis Kellis (10:31.240)

If those variances are large and the differences between the mean of A and the mean of B are

Manolis Kellis (10:36.200)

small, then you have very little power.

Manolis Kellis (10:38.920)

The further the means go apart, that's the effect size, the more power you have, and

Manolis Kellis (10:44.600)

the smaller the standard deviation, the more power you have.

Lex Fridman (10:48.760)

So basically when you're asking, is that sufficiently large, certainly not for everything, but we

Manolis Kellis (10:54.440)

already have enough power for many of the stronger effects in the more tight distributions.

Lex Fridman (11:01.240)

So that's the hopeful message that there exists parts of the genome that have a strong effect

Manolis Kellis (11:09.840)

that has a small variance.

Lex Fridman (11:13.200)

That's exactly right.

Manolis Kellis (11:14.200)

Unfortunately, those perturbations are the basis of disease in many cases.

Lex Fridman (11:18.400)

So it's not a hopeful message.

Manolis Kellis (11:20.800)

Sometimes it's a terrible message.

Manolis Kellis (11:22.720)

It's basically, well, some people are sick, but if we can figure out what are these contributors

Manolis Kellis (11:27.880)

to sickness, we can then help make them better and help many other people better who don't

Lex Fridman (11:32.760)

carry that exact mutation, but who carry mutations on the same pathways.

Lex Fridman (11:38.960)

And that's what we like to call the allelic series of a gene.

Manolis Kellis (11:42.800)

You basically have many perturbations of the same gene in different people, each with a

Manolis Kellis (11:49.580)

different frequency in the human population and each with a different effect on the individual

Lex Fridman (11:55.200)

that carries them.

Lex Fridman (11:56.200)

So you said in the past there would be these small experiments on perturbations and animal

Lex Fridman (12:03.000)

models.

Lex Fridman (12:04.000)

What does this puzzle solving process look like today?

Lex Fridman (12:08.400)

So we basically have something like 7 billion people in the planet and every one of them

Manolis Kellis (12:13.180)

carries something like 6 million mutations.

Manolis Kellis (12:16.760)

You basically have an enormous matrix of genotype by phenotype by systematically measuring the

Manolis Kellis (12:25.000)

phenotype of these individuals.

Lex Fridman (12:27.960)

And the traditional way of measuring this phenotype has been to look at one trait at

Manolis Kellis (12:32.640)

a time.

Manolis Kellis (12:33.700)

You would gather families and you would sort of paint the pedigrees of a strong effect,

Lex Fridman (12:40.160)

what we like to call Mendelian mutation, so a mutation that gets transmitted in a dominant

Manolis Kellis (12:47.320)

or a recessive, but strong effect form where basically one locus plays a very big role

Manolis Kellis (12:53.240)

in that disease.

Lex Fridman (12:54.240)

And you could then look at carriers versus non carriers in one family, carriers versus

Manolis Kellis (12:59.560)

non carriers in another family and do that for hundreds, sometimes thousands of families

Lex Fridman (13:04.480)

and then trace these inheritance patterns and then figure out what is the gene that

Manolis Kellis (13:08.360)

plays that role.

Lex Fridman (13:09.500)

Is this the matrix that you're showing in talks or lectures?

Lex Fridman (13:14.420)

So that matrix is the input to those stuff that I show in talks.

Lex Fridman (13:21.000)

So basically that matrix has traditionally been strong effect genes.

Lex Fridman (13:24.980)

What the matrix looks like now is instead of pedigrees, instead of families, you basically

Manolis Kellis (13:29.880)

have thousands and sometimes hundreds of thousands of unrelated individuals, each with all of

Manolis Kellis (13:36.720)

their genetic variants and each with their phenotype, for example, height or lipids or,

Lex Fridman (13:43.520)

you know, whether they're sick or not for a particular trait.

Manolis Kellis (13:48.080)

That has been the modern view instead of going to families, going to unrelated individuals

Lex Fridman (13:53.080)

with one phenotype at a time.

Lex Fridman (13:55.760)

And what we're doing now as we're maturing in all of these sciences is that we're doing

Manolis Kellis (14:00.960)

this in the context of large medical systems or enormous cohorts that are very well phenotyped

Manolis Kellis (14:07.960)

across hundreds of phenotypes, sometimes with our complete electronic health record.

Lex Fridman (14:13.720)

So you can now start relating not just one gene segregating one family, not just thousands

Manolis Kellis (14:19.640)

of variants segregating with one phenotype, but now you can do millions of variants versus

Lex Fridman (14:25.080)

hundreds of phenotypes.

Lex Fridman (14:27.120)

And as a computer scientist, I mean, deconvolving that matrix, partitioning it into the layers

Manolis Kellis (14:33.880)

of biology that are associated with every one of these elements is a dream come true.

Manolis Kellis (14:40.160)

It's like the world's greatest puzzle.

Lex Fridman (14:42.840)

And you can now solve that puzzle by throwing in more and more knowledge about the function

Manolis Kellis (14:50.120)

of different genomic regions and how these functions are changed across tissues and in

Lex Fridman (14:56.520)

the context of disease.

Lex Fridman (14:58.100)

And that's what my group and many other groups are doing.

Manolis Kellis (15:00.720)

We're trying to systematically relate this genetic variation with molecular variation

Manolis Kellis (15:05.760)

at the expression level of the genes, at the epigenomic level of the gene regulatory circuitry,

Lex Fridman (15:12.700)

and at the cellular level of what are the functions that are happening in those cells,

Manolis Kellis (15:17.020)

at the single cell level using single cell profiling, and then relate all that vast amount

Manolis Kellis (15:22.340)

of knowledge computationally with the thousands of traits that each of these of thousands

Manolis Kellis (15:29.160)

of variants are perturbing.

Lex Fridman (15:30.800)

I mean, this is something we talked about, I think last time.

Lex Fridman (15:34.280)

So there's these effects at different levels that happen.

Manolis Kellis (15:36.480)

You said at a single cell level, you're trying to see things that happen due to certain perturbations.

Lex Fridman (15:42.800)

And then it's not just like a puzzle of perturbation and disease.

Manolis Kellis (15:49.560)

It's perturbation then effect at a cellular level, then at an organ level, a body, like,

Lex Fridman (15:57.660)

how do you disassemble this into like what your group is working on?

Lex Fridman (16:02.760)

You're basically taking a bunch of the hard problems in the space.

Lex Fridman (16:06.560)

How do you break apart a difficult disease and break it apart into problems that you,

Lex Fridman (16:13.520)

into puzzles that you can now start solving?

Lex Fridman (16:15.520)

So there's a struggle here.

Manolis Kellis (16:17.380)

Super scientists love hard puzzles and they're like, oh, I want to build a method that just

Manolis Kellis (16:22.120)

deconvolves the whole thing computationally.

Lex Fridman (16:24.920)

And that's very tempting and it's very appealing, but biologists just like to decouple that

Manolis Kellis (16:31.640)

complexity experimentally, to just like peel off layers of complexity experimentally.

Lex Fridman (16:36.080)

And that's what many of these modern tools that my group and others have both developed

Lex Fridman (16:40.380)

and used.

Manolis Kellis (16:41.600)

The fact that we can now figure out tricks for peeling off these layers of complexity

Manolis Kellis (16:46.760)

by testing one cell type at a time or by testing one cell at a time.

Lex Fridman (16:53.080)

And you could basically say, what is the effect of these genetic variants associated with

Lex Fridman (16:56.380)

Alzheimer's on human brain?

Manolis Kellis (16:59.360)

Human brain sounds like, oh, it's an organ, of course, just go one organ at a time.

Lex Fridman (17:04.320)

But human brain has of course, dozens of different brain regions and within each of these brain

Manolis Kellis (17:09.500)

regions, dozens of different cell types and every single type of neuron, every single

Manolis Kellis (17:15.080)

type of glial cell between astrocytes, oligodendrocytes, microglia, between all of the neural cells

Lex Fridman (17:24.160)

and the vascular cells and the immune cells that are co inhabiting the brain between the

Manolis Kellis (17:29.880)

different types of excitatory and inhibitory neurons that are sort of interacting with

Lex Fridman (17:34.440)

each other between different layers of neurons in the cortical layers.

Manolis Kellis (17:39.300)

Every single one of these has a different type of function to play in cognition, in

Manolis Kellis (17:47.920)

interaction with the environment, in maintenance of the brain, in energetic needs, in feeding

Manolis Kellis (17:55.280)

the brain with blood, with oxygen, in clearing out the debris that are resulting from the

Lex Fridman (18:01.680)

super high energy production of cognition in humans.

Lex Fridman (18:06.940)

So all of these things are basically potentially deconvolvable computationally, but experimentally,

Manolis Kellis (18:17.040)

you can just do single cell profiling of dozens of regions of the brain across hundreds of

Manolis Kellis (18:21.640)

individuals across millions of cells.

Lex Fridman (18:24.640)

And then now you have pieces of the puzzle that you can then put back together to understand

Manolis Kellis (18:31.440)

that complexity.

Manolis Kellis (18:32.440)

I mean, first of all, the cells in the human brain are the most, maybe I'm romanticizing

Manolis Kellis (18:39.400)

it, but cognition seems to be very complicated.

Lex Fridman (18:42.520)

So separating into the function, breaking Alzheimer's down to the cellular level seems

Manolis Kellis (18:53.520)

very challenging.

Manolis Kellis (18:56.340)

Is that basically you're trying to find a way that some perturbation in the genome results

Manolis Kellis (19:05.200)

in some obvious major dysfunction in the cell.

Lex Fridman (19:11.920)

You're trying to find something like that.

Manolis Kellis (19:14.400)

Exactly.

Lex Fridman (19:15.400)

So what does human genetics do?

Manolis Kellis (19:17.120)

Human genetics basically looks at the whole path from genetic variation all the way to

Lex Fridman (19:21.640)

disease.

Lex Fridman (19:22.640)

So human genetics has basically taken thousands of Alzheimer's cases and thousands of controls

Lex Fridman (19:31.640)

matched for age, for sex, for environmental backgrounds and so on and so forth.

Lex Fridman (19:38.440)

And then looked at that map where you're asking, what are the individual genetic perturbations

Lex Fridman (19:44.500)

and how are they related to all the way to Alzheimer's disease?

Lex Fridman (19:48.520)

And that has actually been quite successful.

Lex Fridman (19:51.280)

So we now have more than 27 different loci, these are genomic regions that are associated

Manolis Kellis (19:57.680)

with Alzheimer's at these end to end level.

Manolis Kellis (1:00:04.120)

We have an Alzheimer's and neurodegeneration focus on Huntington's disease, ALS and, you

Manolis Kellis (1:00:10.280)

know, AD related disorders like frontotemporal dementia and Lewy body dementia.

Lex Fridman (1:00:14.460)

And of course, a huge focus on Alzheimer's.

Manolis Kellis (1:00:16.740)

We have a metabolic focus on the role of exercise and diets and sort of how they're impacting

Lex Fridman (1:00:23.320)

metabolic organs across the body and across many different tissues.

Lex Fridman (1:00:29.120)

And all of them are interfacing with the circuitry.

Lex Fridman (1:00:34.100)

And the reason for that is another computer science principle of eat your own dog food.

Manolis Kellis (1:00:42.180)

If everybody ate their own dog food, dog food would taste a lot better.

Manolis Kellis (1:00:47.760)

The reason why Microsoft Excel and Word and PowerPoint was so important and so successful

Manolis Kellis (1:00:55.080)

is because the employees that were working on them, were using them for their day to

Lex Fridman (1:01:00.000)

day tasks.

Manolis Kellis (1:01:01.500)

You can't just simply build a circuitry and say, here it is guys, take the circuitry,

Lex Fridman (1:01:06.120)

we're done without being the users of that circuitry because you then go back.

Lex Fridman (1:01:11.440)

And because we span the whole spectrum from profiling the epigenome, using comparative

Manolis Kellis (1:01:16.740)

genomics, finding the important nucleotides in the genome, building the basic functional

Manolis Kellis (1:01:21.220)

map of what are the genes in the human genome, what are the gene regulatory elements of the

Lex Fridman (1:01:26.800)

human genome.

Manolis Kellis (1:01:27.800)

I mean, over the years we've written a series of papers on how do you find human genes in

Lex Fridman (1:01:31.720)

the first place using comparative genomics?

Lex Fridman (1:01:34.080)

How do you find the motifs that are the building blocks of gene regulation using comparative

Lex Fridman (1:01:38.840)

genomics?

Lex Fridman (1:01:39.840)

And how do you then find how these motifs come together and act in specific tissues

Lex Fridman (1:01:44.800)

using epigenomics?

Lex Fridman (1:01:46.280)

How do you link regulators to enhancers and enhancers to their target genes using epigenomics

Lex Fridman (1:01:53.860)

and regulatory genomics?

Lex Fridman (1:01:55.260)

So through the years we've basically built all this infrastructure for understanding

Lex Fridman (1:02:00.320)

what I like to say, every single nucleotide of the human genome and how it acts in every

Manolis Kellis (1:02:06.900)

one of the major cell types and tissues of the human body.

Lex Fridman (1:02:10.320)

I mean, this is no small task.

Manolis Kellis (1:02:12.040)

This is an enormous task that takes the entire field.

Lex Fridman (1:02:15.540)

And that's something that my group has taken on along with many other groups.

Lex Fridman (1:02:20.720)

And we have also, and that sort of a thing sets my group perhaps apart, we have also

Manolis Kellis (1:02:25.340)

worked with specialists in every one of these disorders to basically further our understanding

Manolis Kellis (1:02:30.640)

all the way down to disease and in some cases collaborating with pharma to go all the way

Manolis Kellis (1:02:35.280)

down to therapeutics because of our deep, deep understanding of that basic circuitry

Lex Fridman (1:02:42.480)

and how it allows us to now improve the circuitry.

Manolis Kellis (1:02:47.600)

Not just treat it as a black box, but basically go and say, okay, we need a better cell type

Manolis Kellis (1:02:51.880)

specific wiring that we now have at the tissue specific level.

Lex Fridman (1:02:56.480)

So we're focusing on that because we're understanding the needs from the disease front.

Lex Fridman (1:03:01.560)

So you have a sense of the entire pipeline, I mean, one, maybe you can indulge me.

Manolis Kellis (1:03:08.040)

One nice question to ask would be, how do you, from the scientific perspective, go from

Manolis Kellis (1:03:14.700)

knowing nothing about the disease to going, you said, to go into the entire pipeline and

Lex Fridman (1:03:22.040)

actually have a drug or a treatment that cures that disease?

Lex Fridman (1:03:26.840)

So that's an enormously long path and an enormously great challenge.

Lex Fridman (1:03:32.840)

And what I'm trying to argue is that it progresses in stages of understanding rather than one

Manolis Kellis (1:03:39.560)

gene at a time.

Manolis Kellis (1:03:40.960)

The traditional view of biology was you have one postdoc working on this gene and another

Manolis Kellis (1:03:45.200)

postdoc working on that gene, and they'll just figure out everything about that gene

Lex Fridman (1:03:50.260)

and that's their job.

Lex Fridman (1:03:52.120)

But we've realized how polygenic the diseases are, so we can't have one postdoc per gene

Lex Fridman (1:03:57.840)

anymore.

Manolis Kellis (1:03:58.840)

We now have to have these cross cutting needs.

Lex Fridman (1:04:04.360)

And I'm going to describe the path to circuitry along those needs.

Lex Fridman (1:04:10.480)

And every single one of these paths, we are now doing in parallel across thousands of

Lex Fridman (1:04:15.600)

genes.

Lex Fridman (1:04:17.000)

So the first step is you have a genetic association, and we talked a little bit about sort of the

Lex Fridman (1:04:23.160)

Mendelian path and the polygenic path to that association.

Lex Fridman (1:04:27.760)

So the Mendelian path was looking through families to basically find gene regions and

Lex Fridman (1:04:33.320)

ultimately genes that are underlying particular disorders.

Manolis Kellis (1:04:36.860)

The polygenic path is basically looking at unrelated individuals in this giant matrix

Manolis Kellis (1:04:43.240)

of genotype by phenotype, and then finding hits where a particular variant impacts disease

Manolis Kellis (1:04:49.200)

all the way to the end.

Lex Fridman (1:04:51.520)

And then we now have a connection, not between a gene and a disease, but between a genetic

Manolis Kellis (1:04:57.960)

region and a disease.

Lex Fridman (1:05:00.200)

And that distinction is not understood by most people.

Lex Fridman (1:05:03.520)

So I'm going to explain it a little bit more.

Lex Fridman (1:05:06.640)

Why do we not have a connection between a gene and a disease, but we have a connection

Lex Fridman (1:05:11.240)

between a genetic region and a disease?

Manolis Kellis (1:05:13.480)

The reason for that is that 93% of genetic variants that are associated with disease

Manolis Kellis (1:05:21.840)

don't impact the protein at all.

Lex Fridman (1:05:27.180)

So if you look at the human genome, there's 20,000 genes, there's 3.2 billion nucleotides.

Manolis Kellis (1:05:33.340)

Only 1.5% of the genome codes for proteins.

Lex Fridman (1:05:40.120)

The other 98.5% does not code for proteins.

Manolis Kellis (1:05:46.160)

If you now look at where are the disease variants located, 93% of them fall in that outside

Lex Fridman (1:05:54.440)

the genes portion.

Manolis Kellis (1:05:55.720)

Of course, genes are enriched, but they're only enriched by a factor of three.

Lex Fridman (1:06:00.600)

That means that still 93% of genetic variants fall outside the proteins.

Lex Fridman (1:06:06.880)

Why is that difficult?

Lex Fridman (1:06:08.200)

Why is that a problem?

Manolis Kellis (1:06:09.480)

The problem is that when a variant falls outside the gene, you don't know what gene is impacted

Lex Fridman (1:06:15.900)

by that variant.

Manolis Kellis (1:06:16.900)

You can't just say, oh, it's near this gene, let's just connect that variant to the gene.

Lex Fridman (1:06:21.160)

And the reason for that is that the genome circuitry is very often long range.

Lex Fridman (1:06:27.880)

So you basically have that genetic variant that could sit in the intron of one gene.

Lex Fridman (1:06:34.880)

An intron is sort of the place between the exons that code for proteins.

Lex Fridman (1:06:38.120)

So proteins are split up into exons and introns and every exon codes for a particular subset

Manolis Kellis (1:06:43.560)

of amino acids and together they're spliced together and then make the final protein.

Lex Fridman (1:06:49.220)

So that genetic variant might be sitting in an intron of a gene.

Manolis Kellis (1:06:51.900)

It's transcribed with the gene, it's processed and then excised, but it might not impact

Manolis Kellis (1:06:56.320)

this gene at all.

Lex Fridman (1:06:57.320)

It might actually impact another gene that's a million nucleotides away.

Lex Fridman (1:07:01.080)

So it's just riding along even though it has nothing to do with this nearby neighborhood.

Lex Fridman (1:07:05.840)

That's exactly right.

Manolis Kellis (1:07:06.840)

Let me give you an example.

Manolis Kellis (1:07:09.600)

The strongest genetic association with obesity was discovered in this FTO gene, fat and obesity

Manolis Kellis (1:07:16.520)

associated gene.

Lex Fridman (1:07:18.400)

So this FTO gene was studied ad nauseum.

Manolis Kellis (1:07:23.780)

People did tons of experiments on it.

Lex Fridman (1:07:26.740)

They figured out that FTO is in fact RNA methylation transferase.

Manolis Kellis (1:07:33.000)

It basically impacts something that we call the epitranscriptome.

Manolis Kellis (1:07:38.880)

Just like the genome can be modified, the transcriptome, the transcript of the genes

Manolis Kellis (1:07:43.520)

can be modified.

Lex Fridman (1:07:44.900)

And we basically said, oh great, that means that epitranscriptomics is hugely involved

Manolis Kellis (1:07:49.320)

in obesity because that gene FTO is clearly where the genetic locus is at.

Manolis Kellis (1:07:56.880)

My group studied FTO in collaboration with a wonderful team led by Melina Klausnitzer.

Lex Fridman (1:08:04.400)

And what we found is that this FTO locus, even though it is as associated with obesity,

Lex Fridman (1:08:11.800)

does not implicate the FTO gene.

Manolis Kellis (1:08:16.680)

The genetic variance, it's in the first intron of the FTO gene, but it controls two genes

Manolis Kellis (1:08:22.840)

IRX3 and IRX5 that are sitting 1.2 million nucleotides away, several genes away.

Manolis Kellis (1:08:32.120)

Oh boy.

Lex Fridman (1:08:33.120)

What am I supposed to feel about that because isn't that like super complicated then?

Lex Fridman (1:08:38.880)

So the way that I was introduced at a conference a few years ago was, and here's Manolis Kellis

Lex Fridman (1:08:43.640)

who wrote the most depressing paper of 2015.

Lex Fridman (1:08:48.720)

And the reason for that is that the entire pharmaceutical industry was so comfortable

Lex Fridman (1:08:52.080)

that there was a single gene in that locus.

Manolis Kellis (1:08:56.120)

Because in some loci, you basically have three dozen genes that are all sitting in the same

Lex Fridman (1:08:59.580)

region of association and you're like, oh gosh, which ones of those is it?

Lex Fridman (1:09:04.060)

But even that question of which ones of those is it is making the assumption that it is

Manolis Kellis (1:09:08.120)

one of those as opposed to some random gene just far, far away, which is what our paper

Manolis Kellis (1:09:13.680)

showed.

Lex Fridman (1:09:14.680)

So basically what our paper showed is that you can't ignore the circuitry.

Manolis Kellis (1:09:19.040)

You have to first figure out the circuitry, all of those long range interactions, how

Manolis Kellis (1:09:23.460)

every genetic variant impacts the expression of every gene in every tissue imaginable across

Manolis Kellis (1:09:28.820)

hundreds of individuals.

Lex Fridman (1:09:30.960)

And then you now have one of the building blocks, not even all of the building blocks

Manolis Kellis (1:09:35.560)

for then going and understanding disease.

Lex Fridman (1:09:41.440)

So embrace the wholeness of the circuitry.

Manolis Kellis (1:09:44.920)

Correct.

Lex Fridman (1:09:45.920)

So back to the question of starting knowing nothing to the disease and going to the treatment.

Lex Fridman (1:09:51.760)

So what are the next steps?

Lex Fridman (1:09:53.520)

So you basically have to first figure out the tissue and then describe how you figure

Manolis Kellis (1:09:57.240)

out the tissue.

Manolis Kellis (1:09:58.240)

You figure out the tissue by taking all of these non coding variants that are sitting

Manolis Kellis (1:10:01.740)

outside proteins and then figuring out what are the epigenomic enrichments.

Lex Fridman (1:10:06.840)

And the reason for that, you know, thankfully is that there is convergence, that the same

Manolis Kellis (1:10:13.860)

processes are impacted in different ways by different loci.

Lex Fridman (1:10:19.440)

And that's a saving grace for our field.

Manolis Kellis (1:10:23.080)

The fact that if I look at hundreds of genetic variants associated with Alzheimer's, they

Lex Fridman (1:10:27.800)

localize in a small number of processes.

Lex Fridman (1:10:31.920)

Can you clarify why that's hopeful?

Lex Fridman (1:10:34.640)

So like they show up in the same exact way in the, in the specific set of processes.

Manolis Kellis (1:10:40.080)

Yeah.

Lex Fridman (1:10:41.080)

So basically there's a small number of biological processes that underlie, or at least that

Manolis Kellis (1:10:45.380)

play the biggest role in every disorder.

Lex Fridman (1:10:48.580)

So in Alzheimer's you basically have, you know, maybe 10 different types of processes.

Manolis Kellis (1:10:54.040)

One of them is lipid metabolism.

Lex Fridman (1:10:56.360)

One of them is immune cell function.

Manolis Kellis (1:10:58.920)

One of them is neuronal energetics.

Lex Fridman (1:11:02.400)

So these are just a small number of processes, but you have multiple lesions, multiple genetic

Manolis Kellis (1:11:07.760)

perturbations that are associated with those processes.

Lex Fridman (1:11:10.980)

So if you look at schizophrenia, it's excitatory neuron function, it's inhibitory neuron function,

Manolis Kellis (1:11:15.800)

it's synaptic pruning, it's calcium signaling and so on and so forth.

Lex Fridman (1:11:18.940)

So when you look at disease genetics, you have one hit here and one hit there and one

Manolis Kellis (1:11:24.840)

hit there and one hit there, completely different parts of the genome.

Lex Fridman (1:11:28.200)

But it turns out all of those hits are calcium signaling proteins.

Manolis Kellis (1:11:31.640)

Oh, cool.

Lex Fridman (1:11:32.640)

You're like, aha.

Manolis Kellis (1:11:34.600)

That means that calcium signaling is important.

Lex Fridman (1:11:37.420)

So those people who are focusing on one doctor at a time cannot possibly see that picture.

Manolis Kellis (1:11:42.640)

You have to become a genomicist.

Manolis Kellis (1:11:44.560)

You have to look at the omics, the om, the holistic picture to understand these enrichments.

Lex Fridman (1:11:51.400)

But you mentioned the convergence thing.

Lex Fridman (1:11:54.080)

The whatever the thing associated with the disease shows up.

Lex Fridman (1:11:58.400)

So let me explain convergence.

Lex Fridman (1:12:00.200)

Convergence is such a beautiful concept.

Lex Fridman (1:12:03.580)

So you basically have these four genes that are converging on calcium signaling.

Lex Fridman (1:12:12.480)

So that basically means that they are acting each in their own way, but together in the

Manolis Kellis (1:12:18.040)

same process.

Lex Fridman (1:12:19.820)

But now in every one of these loci, you have many enhancers controlling each of those genes.

Manolis Kellis (1:12:27.600)

That's another type of convergence where dysregulation of seven different enhancers might all converge

Lex Fridman (1:12:33.280)

on dysregulation of that one gene, which then converges on calcium signaling.

Lex Fridman (1:12:39.280)

And in each one of those enhancers, you might have multiple genetic variants distributed

Lex Fridman (1:12:44.160)

across many different people.

Manolis Kellis (1:12:46.960)

Everyone has their own different mutation.

Lex Fridman (1:12:49.840)

But all of these mutations are impacting that enhancer.

Lex Fridman (1:12:52.880)

And all of these enhancers are impacting that gene.

Lex Fridman (1:12:55.160)

And all of these genes are impacting this pathway.

Lex Fridman (1:12:57.560)

And all these pathways are acting in the same tissue.

Lex Fridman (1:13:00.020)

And all of these tissues are converging together on the same biological process of schizophrenia.

Lex Fridman (1:13:05.280)

And you're saying the saving grace is that that conversion seems to happen for a lot

Lex Fridman (1:13:09.960)

of these diseases.

Manolis Kellis (1:13:11.120)

For all of them.

Manolis Kellis (1:13:12.180)

Basically that for every single disease that we've looked at, we have found an epigenomic

Manolis Kellis (1:13:17.200)

enrichment.

Lex Fridman (1:13:18.500)

How do you do that?

Manolis Kellis (1:13:19.500)

You basically have all of the genetic variants associated with the disorder.

Lex Fridman (1:13:24.040)

And then you're asking for all of the enhancers active in a particular tissue.

Manolis Kellis (1:13:28.080)

For 540 disorders, we've basically found that indeed there is an enrichment.

Lex Fridman (1:13:33.760)

That basically means that there is commonality.

Lex Fridman (1:13:37.060)

And from the commonality, we can just get insights.

Lex Fridman (1:13:40.600)

So to explain in mathematical terms, we're basically building an empirical prior.

Manolis Kellis (1:13:47.120)

We're using a Bayesian approach to basically say, great, all of these variants are equally

Lex Fridman (1:13:52.600)

likely in a particular locus to be important.

Lex Fridman (1:13:57.200)

So in a genetic locus, you basically have a dozen variants that are coinherited.

Manolis Kellis (1:14:02.800)

Because the way that inheritance works in the human genome is through all of these recombination

Manolis Kellis (1:14:07.960)

events during meiosis, you basically have, you know, you inherit maybe three, chromosome

Lex Fridman (1:14:16.120)

three, for example, in your body is inherited from four different parts.

Manolis Kellis (1:14:20.240)

One part comes from your dad, another part comes from your mom, another part comes from

Lex Fridman (1:14:23.840)

your dad, another part comes from your mom.

Lex Fridman (1:14:25.860)

So basically, the way that it, sorry, from your mom's mom.

Lex Fridman (1:14:30.200)

So you basically have one copy that comes from your dad and one copy that comes from

Manolis Kellis (1:14:33.800)

your mom.

Lex Fridman (1:14:34.800)

But that copy that you got from your mom is a mixture of her maternal and her paternal

Manolis Kellis (1:14:39.600)

chromosome.

Lex Fridman (1:14:41.000)

And the copy that you got from your dad is a mixture of his maternal and his paternal

Manolis Kellis (1:14:44.680)

chromosome.

Lex Fridman (1:14:45.680)

So these breakpoints that happen when chromosomes are lining up are basically ensuring through

Manolis Kellis (1:14:53.480)

these crossover events, they're ensuring that every child cell during the process of meiosis,

Manolis Kellis (1:15:02.520)

where you basically have, you know, one spermatozoid that basically couples with one ovule to basically

Manolis Kellis (1:15:08.560)

create one egg to basically create the zygote.

Manolis Kellis (1:15:12.240)

You basically have half of your genome that comes from dad and half your genome that comes

Manolis Kellis (1:15:16.440)

from mom.

Lex Fridman (1:15:17.440)

But in order to line them up, you basically have these crossover events.

Manolis Kellis (1:15:21.040)

These crossover events are basically leading to coinheritance of that entire block coming

Manolis Kellis (1:15:27.880)

from your maternal grandmother and that entire block coming from your maternal grandfather.

Manolis Kellis (1:15:33.920)

Over many generations, these crossover events don't happen randomly.

Manolis Kellis (1:15:38.800)

There's a protein called PRDM9 that basically guides the double stranded breaks and then

Manolis Kellis (1:15:45.720)

leads to these crossovers.

Lex Fridman (1:15:48.320)

And that protein has a particular preference to only a small number of hotspots of recombination,

Manolis Kellis (1:15:54.240)

which then lead to a small number of breaks between these coinheritance patterns.

Lex Fridman (1:15:59.880)

So even though there are 6 million variants, there are 6 million loci, this variation is

Manolis Kellis (1:16:06.720)

inherited in blocks and every one of these blocks has like two dozen genetic variants

Lex Fridman (1:16:12.600)

that are all associated.

Lex Fridman (1:16:13.600)

So in the case of FTO, it wasn't just one variant, it was 89 common variants that were

Lex Fridman (1:16:19.840)

all humongously associated with obesity.

Lex Fridman (1:16:24.320)

Which one of those is the important one?

Lex Fridman (1:16:26.640)

Well, if you look at only one locus, you have no idea.

Lex Fridman (1:16:29.640)

But if you look at many loci, you basically say, aha, all of them are enriching in the

Lex Fridman (1:16:36.880)

same epigenomic map.

Manolis Kellis (1:16:40.080)

In that particular case, it was mesenchymal stem cells.

Lex Fridman (1:16:44.160)

So these are the progenitor cells that give rise to your brown fat and your white fat.

Lex Fridman (1:16:50.560)

Progenitor is like the early on developmental stem cells?

Lex Fridman (1:16:54.020)

So you start from one zygote and that's a totipotent cell type.

Manolis Kellis (1:16:58.120)

It can do anything.

Manolis Kellis (1:17:00.000)

You then, you know, that cell divides, divides, divides, and then every cell division is leading

Manolis Kellis (1:17:08.280)

to specialization where you now have a mesodermal lineage and ectodermal lineage and endodermal

Lex Fridman (1:17:14.880)

lineage that basically leads to different parts of your body.

Manolis Kellis (1:17:19.320)

The ectoderm will basically give rise to your skin, ecto means outside, derm is skin.

Lex Fridman (1:17:25.840)

So ectoderm, but it also gives rise to your neurons and your whole brain.

Lex Fridman (1:17:29.640)

So that's a lot of ectoderm.

Manolis Kellis (1:17:31.600)

Mesoderm gives rise to your internal organs, including the vasculature and you know, your

Manolis Kellis (1:17:36.880)

muscle and stuff like that.

Lex Fridman (1:17:38.440)

So you basically have this progressive differentiation and then if you look further, further down

Manolis Kellis (1:17:45.080)

that lineage, you basically have one lineage that will give rise to both your muscle and

Lex Fridman (1:17:49.700)

your bone, but also your fat.

Lex Fridman (1:17:52.880)

And if you go further down the lineage of your fat, you basically have your white fat

Lex Fridman (1:17:57.720)

cells.

Manolis Kellis (1:17:59.040)

These are the cells that store energy.

Lex Fridman (1:18:01.640)

So when you eat a lot, but you don't exercise too much, there's an excess set of calories,

Manolis Kellis (1:18:06.640)

excess energy.

Lex Fridman (1:18:07.640)

What do you do with those?

Manolis Kellis (1:18:08.640)

You basically create, you spend a lot of that energy to create these high energy molecules,

Lex Fridman (1:18:13.520)

lipids, which you can then burn when you need them on a rainy day.

Lex Fridman (1:18:19.840)

So that leads to obesity if you don't exercise and if you overeat because your body's like,

Lex Fridman (1:18:26.320)

oh great, I have all these calories.

Manolis Kellis (1:18:27.680)

I'm going to store them.

Lex Fridman (1:18:28.680)

Ooh, more calories.

Manolis Kellis (1:18:29.680)

I'm going to store them too.

Lex Fridman (1:18:30.680)

Ooh, more calories.

Lex Fridman (1:18:31.680)

So basically the 42% of European chromosomes have a predisposition to storing fat, which

Manolis Kellis (1:18:40.280)

was selected probably in the food scarcity periods, like basically as we were exiting

Manolis Kellis (1:18:48.880)

Africa before and during the ice ages, there was probably a selection to those individuals

Lex Fridman (1:18:54.240)

who made it North to basically be able to store energy, a lot more energy.

Lex Fridman (1:19:00.880)

So you basically now have this lineage that is deciding whether you want to store energy

Lex Fridman (1:19:07.160)

in your white fat or burn energy in your beige fat.

Manolis Kellis (1:19:11.160)

It turns out that your fat is, you know, like we have such a bad view of fat.

Lex Fridman (1:19:18.680)

Fat is your best friend.

Manolis Kellis (1:19:20.160)

Fat can both store all these excess lipids that would be otherwise circulating through

Lex Fridman (1:19:24.500)

your body and causing damage, but it can also burn calories directly.

Manolis Kellis (1:19:29.900)

If you have too much energy, you can just choose to just burn some of that as heat.

Lex Fridman (1:19:35.760)

So basically when you're cold, you're burning energy to basically warm your body up and

Manolis Kellis (1:19:41.200)

you're burning all these lipids and you're burning all these calories.

Lex Fridman (1:19:44.540)

So what we basically found is that across the board, genetic variants associated with

Manolis Kellis (1:19:50.000)

obesity across many of these regions were all enriched repeatedly in mesenchymal stem

Lex Fridman (1:19:56.520)

cell enhancers.

Lex Fridman (1:19:58.360)

So that gave us a hint as to which of these genetic variants was likely driving this whole

Lex Fridman (1:20:05.120)

association.

Lex Fridman (1:20:06.120)

And we ended up with this one genetic variant called RS1421085.

Lex Fridman (1:20:14.440)

And that genetic variant out of the 89 was the one that we predicted to be causal for

Manolis Kellis (1:20:20.040)

the disease.

Lex Fridman (1:20:21.040)

Wow.

Lex Fridman (1:20:22.040)

So going back to those steps, first step is figure out the relevant tissue based on the

Lex Fridman (1:20:26.240)

global enrichment.

Manolis Kellis (1:20:27.960)

Second step is figure out the causal variant among many variants in this linkage disequilibrium

Manolis Kellis (1:20:34.840)

in this coinherited block between these recombination hotspots, these boundaries of these inherited

Manolis Kellis (1:20:41.160)

blocks.

Lex Fridman (1:20:42.640)

That's the second step.

Manolis Kellis (1:20:43.920)

The third step is once you know that causal variant, try to figure out what is the motif

Lex Fridman (1:20:49.920)

that is disrupted by that causal variant.

Lex Fridman (1:20:52.720)

Basically how does it act?

Manolis Kellis (1:20:54.400)

Variants don't just disrupt elements, they disrupt the binding of specific regulators.

Lex Fridman (1:20:59.520)

So basically the third step there was how do you find the motif that is responsible

Manolis Kellis (1:21:04.440)

like the gene regulatory word, the building block of gene regulation that is responsible

Manolis Kellis (1:21:10.240)

for that dysregulatory event.

Lex Fridman (1:21:12.480)

And the fourth step is finding out what regulator normally binds that motif and is now no longer

Manolis Kellis (1:21:18.280)

able to bind.

Lex Fridman (1:21:19.280)

And then once you have the regulator, can you then try to figure out how to, what after

Lex Fridman (1:21:24.920)

it developed, how to fix it?

Lex Fridman (1:21:27.200)

That's exactly right.

Manolis Kellis (1:21:28.200)

You now know how to intervene.

Manolis Kellis (1:21:30.260)

You have basically a regulator, you have a gene that you can then perturb and you say,

Manolis Kellis (1:21:34.520)

well, maybe that regulator has a global role in obesity.

Lex Fridman (1:21:38.640)

I can perturb the regulator.

Manolis Kellis (1:21:40.360)

Just to clarify, when we say perturb, like on the scale of a human life, can a human

Lex Fridman (1:21:46.760)

being be helped?

Manolis Kellis (1:21:49.000)

Of course.

Lex Fridman (1:21:50.000)

Yeah.

Manolis Kellis (1:21:51.000)

I guess understanding is the first step.

Manolis Kellis (1:21:52.480)

No, no, but perturbed basically means you now develop therapeutics, pharmaceutical therapeutics

Manolis Kellis (1:21:57.480)

against that.

Manolis Kellis (1:21:59.340)

Or you develop other types of intervention that affect the expression of that gene.

Lex Fridman (1:22:03.800)

What do pharmaceutical therapeutics look like when your understanding is on a genetic level?

Lex Fridman (1:22:11.040)

Yeah.

Manolis Kellis (1:22:12.040)

Sorry if it's a dumb question.

Lex Fridman (1:22:13.040)

No, no, no.

Manolis Kellis (1:22:14.040)

It's a brilliant question, but I want to save it for a little bit later when we start talking

Lex Fridman (1:22:16.440)

about therapeutics.

Manolis Kellis (1:22:17.440)

Perfect.

Lex Fridman (1:22:18.440)

So let's talk about the first four steps.

Manolis Kellis (1:22:20.280)

There's two more.

Lex Fridman (1:22:21.600)

So basically the first step is figure out, I mean, the zero step, the starting point

Manolis Kellis (1:22:25.600)

is the genetics.

Lex Fridman (1:22:26.760)

The first step after that is figure out the tissue of action.

Manolis Kellis (1:22:31.100)

The second step is figuring out the nucleotide that is responsible or set of nucleotides.

Manolis Kellis (1:22:36.920)

The third step is figuring out the motif and the upstream regulator, number four.

Lex Fridman (1:22:40.960)

Number five and six is what are the targets?

Lex Fridman (1:22:44.320)

So number five is great.

Manolis Kellis (1:22:45.800)

Now I know the regulator.

Lex Fridman (1:22:47.200)

I know the motif.

Manolis Kellis (1:22:48.200)

I know the tissue and I know the variant.

Lex Fridman (1:22:51.460)

What does it actually do?

Lex Fridman (1:22:53.400)

So you have to now trace it to the biological process and the genes that mediate that biological

Lex Fridman (1:22:59.240)

process.

Lex Fridman (1:23:00.480)

So knowing all of this can now allow you to find the target genes.

Lex Fridman (1:23:05.400)

How?

Manolis Kellis (1:23:06.400)

By basically doing perturbation experiments or by looking at the folding of the epigenome

Manolis Kellis (1:23:13.200)

or by looking at the genetic impact of that genetic variant on the expression of genes.

Lex Fridman (1:23:19.440)

And we use all three.

Lex Fridman (1:23:21.580)

So let me go through them.

Manolis Kellis (1:23:22.800)

Basically one of them is physical links.

Lex Fridman (1:23:26.360)

This is the folding of the genome onto itself.

Lex Fridman (1:23:29.920)

How do you even figure out the folding?

Lex Fridman (1:23:32.200)

It's a little bit of a tangent, but it's a super awesome technology.

Manolis Kellis (1:23:36.760)

Think of the genome as again, this massive packaging that we talked about of taking two

Manolis Kellis (1:23:41.960)

meters worth of DNA and putting it in something that's a million times smaller than two meters

Manolis Kellis (1:23:48.960)

worth of DNA.

Lex Fridman (1:23:49.960)

That's a single cell.

Manolis Kellis (1:23:51.760)

You basically have this massive packaging and this packaging basically leads to the

Manolis Kellis (1:23:56.160)

chromosome being wrapped around in sort of tight, tight ways in ways, however, that are

Manolis Kellis (1:24:02.600)

functionally capable of being reopened and reclosed.

Lex Fridman (1:24:07.080)

So I can then go in and figure out that folding by sort of chopping up the spaghetti soup,

Manolis Kellis (1:24:15.000)

putting glue and ligating the segments that were chopped up but nearby each other, and

Manolis Kellis (1:24:21.000)

then sequencing through these ligation events to figure out that this region of this chromosome,

Manolis Kellis (1:24:26.020)

that region of the chromosome were near each other.

Manolis Kellis (1:24:28.360)

That means they were interacting even though they were far away on the genome itself.

Lex Fridman (1:24:33.560)

So that chopping up, sequencing and reglueing is basically giving you folds of the genome

Lex Fridman (1:24:42.500)

that we call.

Lex Fridman (1:24:43.500)

Sorry, can you backtrack?

Lex Fridman (1:24:44.500)

Of course.

Lex Fridman (1:24:45.500)

How does cutting it help you figure out which ones were close in the original folding?

Lex Fridman (1:24:50.600)

So you have a bowl of noodles.

Manolis Kellis (1:24:53.440)

Go on.

Lex Fridman (1:24:54.760)

And in that bowl of noodles, some noodles are near each other.

Manolis Kellis (1:24:59.480)

Yes.

Lex Fridman (1:25:00.480)

So you throw in a bunch of glue, you basically freeze the noodles in place, throw in a cutter

Manolis Kellis (1:25:06.520)

that chops up the noodles into little pieces.

Manolis Kellis (1:25:10.860)

Now throw in some ligation enzyme that lets those pieces that were free religate near

Manolis Kellis (1:25:18.040)

each other.

Lex Fridman (1:25:19.080)

In some cases, they religate what you had just cut, but that's very rare.

Manolis Kellis (1:25:24.240)

Most of the time they will religate in whatever was proximal.

Manolis Kellis (1:25:30.320)

You now have glued the red noodle that was crossing the blue noodle to each other.

Manolis Kellis (1:25:36.760)

You then reverse the glue, the glue goes away and you just sequence the heck out of it.

Manolis Kellis (1:25:43.020)

Most of the time you'll find red segment with, you know, red segment, but you can specifically

Manolis Kellis (1:25:48.640)

select for ligation events that have happened that were not from the same segment by sort

Manolis Kellis (1:25:52.640)

of marking them in a particular way and then selecting those and then you sequence and

Manolis Kellis (1:25:57.360)

you look for red with blue matches of sort of things that were glued that were not immediate

Lex Fridman (1:26:03.400)

proximal to each other.

Lex Fridman (1:26:05.520)

And that reveals the linking of the blue noodle and the red noodle.

Lex Fridman (1:26:08.640)

You're with me so far?

Manolis Kellis (1:26:09.640)

Yeah.

Lex Fridman (1:26:10.640)

Good.

Lex Fridman (1:26:11.640)

So we've done these experiments.

Lex Fridman (1:26:12.640)

That's the physical.

Manolis Kellis (1:26:13.640)

That's the physical.

Lex Fridman (1:26:14.640)

That's step one of the physical.

Lex Fridman (1:26:15.820)

And what the physical revealed is topologically associated domains, basically big blocks of

Lex Fridman (1:26:20.000)

the genome that are topologically connected together.

Manolis Kellis (1:26:25.040)

That's the physical.

Lex Fridman (1:26:26.300)

The second one is the genetic links.

Manolis Kellis (1:26:30.060)

It basically says across individuals that have different genetic variants, how are their

Lex Fridman (1:26:37.220)

genes expressed differently?

Manolis Kellis (1:26:39.400)

Remember before I was saying that the path between genetics and disease is enormous,

Lex Fridman (1:26:43.080)

but we can break it up to look at the path between genetics and gene expression.

Lex Fridman (1:26:47.520)

So instead of using Alzheimer's as a phenotype, I can now use expression of IRX3 as the phenotype,

Manolis Kellis (1:26:54.480)

expression of gene A. And I can look at all of the humans who contain a G at that location

Lex Fridman (1:27:01.160)

and all the humans that contain a T at that location and basically say, wow, it turns

Manolis Kellis (1:27:05.360)

out that the expression of each gene is higher for the T humans than for the G humans at

Manolis Kellis (1:27:09.480)

that location.

Lex Fridman (1:27:10.660)

So that basically gives me a genetic link between a genetic variant, a locus, a region,

Lex Fridman (1:27:16.560)

and the expression of nearby genes.

Lex Fridman (1:27:19.960)

Good on the genetic link?

Manolis Kellis (1:27:20.960)

I think so.

Lex Fridman (1:27:21.960)

Awesome.

Manolis Kellis (1:27:22.960)

The third genetic link is the activity link.

Lex Fridman (1:27:25.480)

What's an activity link?

Manolis Kellis (1:27:26.480)

It basically says if I look across 833 different epigenomes, whenever this enhancer is active,

Lex Fridman (1:27:34.320)

this gene is active.

Manolis Kellis (1:27:36.040)

That gives me an activity link between this region of the DNA and that gene.

Lex Fridman (1:27:42.340)

And then the fourth one is perturbations where I can go in and blow up that region and see

Lex Fridman (1:27:47.140)

what are the genes that change in expression, or I can go in and over activate that region

Lex Fridman (1:27:51.900)

and see what genes change in expression.

Lex Fridman (1:27:55.120)

So I guess that's similar to activity?

Lex Fridman (1:27:57.240)

Yeah.

Manolis Kellis (1:27:58.240)

Yeah.

Lex Fridman (1:27:59.240)

So that's basically similar to activity.

Manolis Kellis (1:28:00.240)

I agree, but it's causal rather than correlational.

Lex Fridman (1:28:02.760)

Again, I'm a little weird.

Manolis Kellis (1:28:04.960)

No, no, you're 100% on.

Lex Fridman (1:28:07.160)

It's exactly the same as the perturbation where I go in and intervene.

Manolis Kellis (1:28:11.440)

I basically take a bunch of cells.

Lex Fridman (1:28:13.800)

So you know CRISPR, right?

Manolis Kellis (1:28:16.160)

CRISPR is this genome guidance and cutting mechanism.

Lex Fridman (1:28:21.500)

That's what George Church likes to call genome vandalism.

Lex Fridman (1:28:24.680)

So you basically are able to, you can basically take a guide RNA that you put into the CRISPR

Manolis Kellis (1:28:32.720)

system, and the CRISPR system will basically use this guide RNA, scan the genome, find

Manolis Kellis (1:28:38.200)

wherever there's a match, and then cut the genome.

Lex Fridman (1:28:42.560)

So I digress, but it's a bacterial immune defense system.

Lex Fridman (1:28:48.000)

So basically bacteria are constantly attacked by viruses, but sometimes they win against

Lex Fridman (1:28:54.280)

the viruses and they chop up these viruses.

Lex Fridman (1:28:56.960)

And remember as a trophy inside their genome, they have these loci, these CRISPR loci that

Lex Fridman (1:29:02.800)

basically stands for clustered repeats, interspersed, et cetera.

Lex Fridman (1:29:06.400)

So basically it's an interspersed repeats structure where basically you have a set of

Manolis Kellis (1:29:11.900)

repetitive regions and then interspersed where these variable segments that were basically

Manolis Kellis (1:29:17.400)

matching viruses.

Lex Fridman (1:29:19.600)

So when this was first discovered, it was basically hypothesized that this is probably

Manolis Kellis (1:29:24.240)

a bacterial immune system that remembers the trophies of the viruses that managed to kill.

Lex Fridman (1:29:30.360)

And then the bacteria pass on, you know, they sort of do lateral transfer of DNA and they

Manolis Kellis (1:29:34.720)

pass on these memories so that the next bacterium says, Ooh, you killed that guy.

Lex Fridman (1:29:39.120)

When that guy shows up again, I will recognize him.

Lex Fridman (1:29:41.700)

And the CRISPR system was basically evolved as a bacterial adaptive immune response to

Lex Fridman (1:29:47.320)

sense foreigners that should not belong and to just go and cut their genome.

Lex Fridman (1:29:52.560)

So it's an RNA guided RNA cutting enzyme or an RNA guided DNA cutting enzyme.

Lex Fridman (1:30:00.280)

So there's different systems.

Manolis Kellis (1:30:02.240)

Some of them cut DNA, some of them cut RNA, but all of them remember this sort of viral

Lex Fridman (1:30:08.620)

attack.

Lex Fridman (1:30:10.660)

So what we have done now as a field is, you know, through the work of, you know, Jennifer

Manolis Kellis (1:30:15.920)

Donne, Manuel Carpentier, Feng Zhang and many others is coopted that system of bacterial

Manolis Kellis (1:30:23.240)

immune defense as a way to cut genomes.

Manolis Kellis (1:30:26.520)

You basically have this guiding system that allows you to use an RNA guide to bring enzymes

Manolis Kellis (1:30:35.280)

to cut DNA at a particular locus.

Lex Fridman (1:30:37.760)

That's so fascinating.

Lex Fridman (1:30:39.240)

So this is like already a natural mechanism, a natural tool for cutting those useful as

Lex Fridman (1:30:45.600)

particular context.

Lex Fridman (1:30:46.600)

And we're like, well, we can use that thing to actually, it's a nice tool that's already

Lex Fridman (1:30:51.160)

in the body.

Manolis Kellis (1:30:52.160)

Yeah.

Lex Fridman (1:30:53.160)

Yeah.

Manolis Kellis (1:30:54.160)

It's not in our body.

Lex Fridman (1:30:55.160)

It's in the bacterial body.

Manolis Kellis (1:30:56.160)

It was discovered by the yogurt industry.

Manolis Kellis (1:30:59.320)

They were trying to make better yogurts and they were trying to make their bacteria in

Manolis Kellis (1:31:03.640)

their yogurt cultures more resilient to viruses.

Lex Fridman (1:31:08.400)

And they were studying bacteria and they found that, wow, this CRISPR system is awesome.

Manolis Kellis (1:31:12.480)

It allows you to defend against that.

Lex Fridman (1:31:14.820)

And then it was coopted in mammalian systems that don't use anything like that as a targeting

Manolis Kellis (1:31:20.600)

way to basically bring these DNA cutting enzymes to any locus in the genome.

Lex Fridman (1:31:25.800)

Why would you want to cut DNA to do anything?

Manolis Kellis (1:31:29.620)

The reason is that our DNA has a DNA repair mechanism where if a region of the genome

Manolis Kellis (1:31:35.040)

gets randomly cut, you will basically scan the genome for anything that matches and sort

Manolis Kellis (1:31:40.520)

of use it by homology.

Lex Fridman (1:31:43.480)

So the reason why we're deployed is because we now have a spare copy.

Manolis Kellis (1:31:47.240)

As soon as my mom's copy is deactivated, I can use my dad's copy.

Lex Fridman (1:31:50.640)

And somewhere else, if my dad's copy is deactivated, I can use my mom's copy to repair it.

Lex Fridman (1:31:55.240)

So this is called homologous based repair.

Lex Fridman (1:31:59.720)

So all you have to do is the cutting and you don't have to do the fixing.

Manolis Kellis (1:32:04.080)

That's exactly right.

Lex Fridman (1:32:05.080)

You don't have to do the fixing.

Manolis Kellis (1:32:06.080)

Because it's already built in.

Lex Fridman (1:32:07.320)

That's exactly right.

Lex Fridman (1:32:08.560)

But the fixing can be coopted by throwing in a bunch of homologous segments that instead

Lex Fridman (1:32:14.720)

of having your dad's version, have whatever other version you'd like to use.

Lex Fridman (1:32:19.960)

So you then control the fixing by throwing in a bunch of other stuff.

Lex Fridman (1:32:24.040)

That's exactly right.

Lex Fridman (1:32:25.040)

And that's how you do genome editing.

Lex Fridman (1:32:26.440)

So that's what CRISPR is.

Manolis Kellis (1:32:27.880)

That's what CRISPR is.

Lex Fridman (1:32:28.880)

In popular culture, people use the term.

Manolis Kellis (1:32:30.840)

I've never, wow, that's brilliant.

Lex Fridman (1:32:32.640)

So CRISPR is genome vandalism followed by a bunch of band aids that have the sequence

Manolis Kellis (1:32:39.080)

that you'd like.

Lex Fridman (1:32:40.080)

And you could control the choices of band aids.

Manolis Kellis (1:32:43.000)

Correct.

Lex Fridman (1:32:44.000)

And of course there's new generations of CRISPR.

Manolis Kellis (1:32:46.360)

There's something that's called prime editing that was sort of very, very much in the press

Manolis Kellis (1:32:50.880)

recently that basically instead of sort of making a double stranded break, which again

Manolis Kellis (1:32:55.360)

is genome vandalism, you basically make a single stranded break.

Manolis Kellis (1:33:00.820)

You basically just nick one of the two strands, enabling you to sort of peel off without sort

Manolis Kellis (1:33:06.640)

of completely breaking it up and then repair it locally using a guide that is coupled to

Lex Fridman (1:33:13.280)

your initial RNA that took you to that location.

Lex Fridman (1:33:18.600)

Dumb question, but is CRISPR as awesome and cool as it sounds?

Manolis Kellis (1:33:24.000)

I mean, technically speaking, in terms of like as a tool for manipulating our genetics

Manolis Kellis (1:33:31.820)

in the positive meaning of the word manipulating, or is there downsides, drawbacks in this whole

Lex Fridman (1:33:39.040)

context of therapeutics that we're talking about or understanding and so on?

Lex Fridman (1:33:42.920)

So when I teach my students about CRISPR, I show them articles with the headline, genome

Lex Fridman (1:33:50.040)

editing tool revolutionizes biology.

Lex Fridman (1:33:53.120)

And then I show them the date of these articles and they're 2004, like five years before CRISPR

Lex Fridman (1:33:58.360)

was invented.

Lex Fridman (1:33:59.760)

And the reason is that they're not talking about CRISPR.

Manolis Kellis (1:34:02.360)

They're talking about zinc finger enzymes that are another way to bring these cutters

Manolis Kellis (1:34:07.520)

to the genome.

Manolis Kellis (1:34:09.040)

It's a very difficult way of sort of designing the right set of zinc finger proteins, the

Manolis Kellis (1:34:13.880)

right set of amino acids that will now target a particular long stretch of DNA because for

Manolis Kellis (1:34:20.280)

every location that you want to target, you need to design a particular regulator, a particular

Manolis Kellis (1:34:25.760)

protein that will match that region well.

Manolis Kellis (1:34:28.800)

There's another technology called talons, which are basically just a different way of

Manolis Kellis (1:34:35.240)

using proteins to sort of guide these cutters to a particular location of the genome.

Manolis Kellis (1:34:41.440)

These require a massive team of engineers, of biological engineers to basically design

Manolis Kellis (1:34:46.520)

a set of amino acids that will target a particular sequence of your genome.

Manolis Kellis (1:34:51.480)

The reason why CRISPR is amazingly, awesomely revolutionary is because instead of having

Manolis Kellis (1:34:57.080)

this team of engineers design a new set of proteins for every locus that you want to

Lex Fridman (1:35:02.200)

target, you just type it in your computer and you just synthesize an RNA guide.

Manolis Kellis (1:35:07.680)

The beauty of CRISPR is not the cutting, it's not the fixing.

Lex Fridman (1:35:11.100)

All of that was there before.

Manolis Kellis (1:35:12.880)

It's the guiding, and the only thing that changes is that it makes the guiding easier

Manolis Kellis (1:35:17.880)

by sort of just typing in the RNA sequence, which then allows the system to sort of scan

Manolis Kellis (1:35:23.940)

the DNA to find that.

Lex Fridman (1:35:25.880)

So the coding, the engineering of the cutter is easier in terms of SP.

Manolis Kellis (1:35:32.280)

That's kind of similar to the story of deep learning versus old school machine learning.

Lex Fridman (1:35:37.200)

Some of the challenging parts are automated.

Lex Fridman (1:35:41.080)

But CRISPR is just one cutting technology, and then that's part of the challenges and

Manolis Kellis (1:35:47.180)

exciting opportunities of the field is to design different cutting technologies.

Lex Fridman (1:35:53.020)

So now this was a big parenthesis on CRISPR, but now when we were talking about perturbations,

Manolis Kellis (1:36:00.840)

you basically now have the ability to not just look at correlation between enhancers

Lex Fridman (1:36:04.720)

and genes, but actually go and either destroy that enhancer and see if the gene changes

Manolis Kellis (1:36:10.760)

in expression, or you can use the CRISPR targeting system to bring in not vandalism and cutting,

Lex Fridman (1:36:20.000)

but you can couple the CRISPR system with, and the CRISPR system is called usually CRISPR

Lex Fridman (1:36:26.720)

Cas9 because Cas9 is the protein that will then come and cut.

Lex Fridman (1:36:30.920)

But there's a version of that protein called dead Cas9 where the cutting part is deactivated.

Lex Fridman (1:36:36.760)

So you basically use the dead Cas9 to bring in an activator or to bring in a repressor.

Lex Fridman (1:36:45.040)

So you can now ask, is this enhancer changing that gene by taking this modified CRISPR,

Manolis Kellis (1:36:51.920)

which is already modified from the bacteria to be used in humans, that you can now modify

Manolis Kellis (1:36:55.560)

the Cas9 to be dead Cas9, and you can now further modify to bring in a regulator, and

Manolis Kellis (1:37:01.120)

you can basically turn on or turn off that enhancer and then see what is the impact on

Manolis Kellis (1:37:05.000)

that gene.

Lex Fridman (1:37:06.620)

So these are the four ways of linking the locus to the target gene, and that's step

Manolis Kellis (1:37:11.840)

number five.

Manolis Kellis (1:37:14.240)

Step number five is find the target gene, and step number six is what the heck does

Lex Fridman (1:37:17.960)

that gene do?

Manolis Kellis (1:37:19.560)

You basically now go and manipulate that gene to basically see what are the processes that

Manolis Kellis (1:37:25.840)

change, and you can basically ask, well, in this particular case, in the FTO locus, we

Manolis Kellis (1:37:32.400)

found mesenchymal stem cells that are the progenitors of white fat and brown fat or

Manolis Kellis (1:37:38.160)

beige fat.

Lex Fridman (1:37:39.580)

We found the RS1421085 nucleotide variant as the causal variant.

Manolis Kellis (1:37:44.880)

We found this large enhancer, this master regulator.

Manolis Kellis (1:37:49.720)

I like to call it OB1 for obesity one, like the strongest enhancer associated with it,

Lex Fridman (1:37:55.720)

and OB1 was kind of chubby as the actor.

Lex Fridman (1:37:57.120)

I don't know if you remember him.

Lex Fridman (1:38:01.120)

So you basically are using this Jedi mind trick to basically find out the location of

Manolis Kellis (1:38:07.320)

the genome that is responsible, the enhancer that harbors it, the motif, the upstream regulator,

Manolis Kellis (1:38:14.120)

which is ARID5B for AT rich interacting domain 5B.

Lex Fridman (1:38:18.200)

That's a protein that sort of comes and binds normally.

Manolis Kellis (1:38:21.040)

That protein is normally a repressor.

Manolis Kellis (1:38:23.220)

It represses this super enhancer, this massive 12,000 nucleotide master regulatory control

Manolis Kellis (1:38:28.520)

gene, and it turns off IRX3, which is a gene that's 600,000 nucleotides away, and IRX5,

Lex Fridman (1:38:36.120)

which is 1.2 million nucleotides away.

Lex Fridman (1:38:38.480)

So those things.

Lex Fridman (1:38:39.480)

And what's the effect of turning them off?

Manolis Kellis (1:38:40.760)

That's exactly the next question.

Lex Fridman (1:38:42.320)

So step six is what do these genes actually do?

Lex Fridman (1:38:45.520)

So we then ask, what does RX3 and RX5 do?

Manolis Kellis (1:38:48.640)

The first thing we did is look across individuals for individuals that had higher expression

Manolis Kellis (1:38:52.940)

of RX3 or lower expression RX3.

Lex Fridman (1:38:55.520)

And then we looked at the expression of all of the other genes in the genome.

Lex Fridman (1:38:58.960)

And we looked for simply correlation.

Lex Fridman (1:39:01.580)

And we found that RX3 and RX5 were both correlated positively with lipid metabolism and negatively

Manolis Kellis (1:39:09.820)

with mitochondrial biogenesis.

Lex Fridman (1:39:11.800)

You're like, what the heck does that mean?

Lex Fridman (1:39:16.400)

Does this sound related to obesity?

Manolis Kellis (1:39:18.120)

Not at all superficially, but lipid metabolism should, because lipids is these high and

Manolis Kellis (1:39:25.500)

energy molecules that basically store fat.

Lex Fridman (1:39:28.560)

So RX3 and RX5 are negatively correlated with lipid metabolism.

Lex Fridman (1:39:33.760)

So that basically means that when they turn on, positively, when they turn on, they turn

Lex Fridman (1:39:39.000)

on lipid metabolism.

Lex Fridman (1:39:41.280)

And they're negatively correlated with mitochondrial biogenesis.

Lex Fridman (1:39:45.920)

What do mitochondria do in this whole process?

Lex Fridman (1:39:49.160)

Again, small parenthesis, what are mitochondria?

Lex Fridman (1:39:53.280)

Mitochondria are little organelles.

Manolis Kellis (1:39:56.360)

They arose, they only are found in eukaryotes.

Lex Fridman (1:40:01.120)

U means good, karyote means nucleus.

Lex Fridman (1:40:04.000)

So truly like a true nucleus.

Lex Fridman (1:40:05.920)

So eukaryotes have a nucleus.

Manolis Kellis (1:40:07.880)

Prokaryotes are before the nucleus.

Lex Fridman (1:40:09.960)

They don't have a nucleus.

Lex Fridman (1:40:11.280)

So eukaryotes have a nucleus, compartmentalization.

Lex Fridman (1:40:16.840)

Eukaryotes have also organelles.

Manolis Kellis (1:40:19.680)

Some eukaryotes have chloroplasts.

Lex Fridman (1:40:22.800)

These are the plants, they photosynthesize.

Manolis Kellis (1:40:26.480)

Some other eukaryotes like us have another type of organelle called mitochondria.

Lex Fridman (1:40:33.480)

These arose from an ancient species that we engulfed.

Manolis Kellis (1:40:40.360)

This is an endosymbiosis event.

Lex Fridman (1:40:44.360)

Symbiosis bio means life, sim means together.

Lex Fridman (1:40:47.320)

So symbiotes are things that live together.

Manolis Kellis (1:40:50.800)

Symbiosis endo means inside, so endosymbiosis means you live together holding the other

Manolis Kellis (1:40:54.240)

one inside you.

Lex Fridman (1:40:56.120)

So the pre eukaryotes engulfed an organism that was very good at energy production and

Manolis Kellis (1:41:07.240)

that organism eventually shed most of its genome to now have only 13 genes in the mitochondrial

Manolis Kellis (1:41:14.200)

genome and those 13 genes are all involved in energy production, the electron transport

Manolis Kellis (1:41:22.400)

chain.

Lex Fridman (1:41:23.400)

So basically electrons are these massive super energy rich molecules.

Manolis Kellis (1:41:28.560)

We basically have these organelles that produce energy and when your muscle exercises, you

Lex Fridman (1:41:35.760)

basically multiply your mitochondria.

Manolis Kellis (1:41:37.800)

You basically sort of, you know, use more and more mitochondria and that's how you get

Lex Fridman (1:41:42.960)

beefed up.

Lex Fridman (1:41:43.960)

So basically the muscle sort of learns how to generate more energy.

Lex Fridman (1:41:47.840)

So basically every single time your muscles will, you know, overnight regenerate and sort

Manolis Kellis (1:41:51.680)

of become stronger and amplify their mitochondria and so forth.

Lex Fridman (1:41:55.240)

So what does mitochondria do?

Manolis Kellis (1:41:56.480)

The mitochondria use energy to sort of do any kind of task.

Lex Fridman (1:42:02.200)

When you're thinking, you're using energy.

Manolis Kellis (1:42:05.000)

This energy comes from mitochondria.

Lex Fridman (1:42:06.960)

Your neurons have mitochondria all over the place.

Manolis Kellis (1:42:10.040)

Basically this mitochondria can multiply as organelles and they can be spread along the

Lex Fridman (1:42:13.340)

body of your muscle.

Manolis Kellis (1:42:15.040)

Some of your muscle cells have actually multiple nuclei, they're polynucleated, but they also

Manolis Kellis (1:42:18.840)

have multiple mitochondria to basically deal with the fact that your muscle is enormous.

Manolis Kellis (1:42:24.380)

You can sort of span these super, super long length and you need energy throughout the

Lex Fridman (1:42:28.040)

length of your muscle.

Lex Fridman (1:42:29.360)

So that's why you have mitochondria throughout the length and you also need transcription

Lex Fridman (1:42:32.340)

through the length so you have multiple nuclei as well.

Lex Fridman (1:42:35.080)

So these two processes, lipids store energy, what do mitochondria do?

Lex Fridman (1:42:42.060)

So there's a process known as thermogenesis.

Manolis Kellis (1:42:46.040)

Thermal heat, genesis generation.

Lex Fridman (1:42:48.520)

Thermogenesis is the generation of heat.

Lex Fridman (1:42:50.600)

Remember that bathtub with the in and out?

Lex Fridman (1:42:55.160)

That's the equation that everybody's focused on.

Lex Fridman (1:42:57.160)

So how much energy do you consume?

Lex Fridman (1:42:58.860)

How much energy do you burn?

Lex Fridman (1:43:01.000)

But in every thermodynamic system, there's three parts to the equation.

Lex Fridman (1:43:06.060)

There's energy in, energy out, and energy lost.

Manolis Kellis (1:43:10.900)

Any machine has loss of energy.

Lex Fridman (1:43:14.680)

How do you lose energy?

Manolis Kellis (1:43:15.720)

You emanate heat.

Lex Fridman (1:43:17.600)

So heat is energy loss.

Lex Fridman (1:43:20.000)

So there's...

Lex Fridman (1:43:24.760)

Which is where the thermogenesis comes in.

Manolis Kellis (1:43:26.600)

Thermogenesis is actually a regulatory process that modulates the third component of the

Lex Fridman (1:43:32.240)

thermodynamic equation.

Manolis Kellis (1:43:34.060)

You can basically control thermogenesis explicitly.

Lex Fridman (1:43:37.240)

You can turn on and turn off thermogenesis.

Lex Fridman (1:43:39.080)

And that's where the mitochondria comes into play.

Lex Fridman (1:43:41.400)

Exactly.

Lex Fridman (1:43:42.400)

So Irix3 and RX5 turn out to be the master regulators of a process of thermogenesis versus

Lex Fridman (1:43:49.600)

lipogenesis generation of fat.

Lex Fridman (1:43:52.360)

So Irix3 and RX5 in most people burn heat, burn calories as heat.

Lex Fridman (1:43:58.720)

So when you eat too much, just burn it off in your fat cells.

Lex Fridman (1:44:02.720)

So that bathtub has basically a sort of dissipation knob that most people are able to turn on.

Manolis Kellis (1:44:11.140)

I am unable to turn that on because I am a homozygous carrier for the mutation that changes

Manolis Kellis (1:44:17.720)

a T into a C in the RS1421085 allele and locus, a SNP.

Lex Fridman (1:44:24.560)

I have the risk allele twice from my mom and from my dad.

Lex Fridman (1:44:28.320)

So I'm unable to thermogenize.

Manolis Kellis (1:44:31.880)

I'm unable to turn on thermogenesis through Irix3 and RX5 because the regulator that normally

Manolis Kellis (1:44:37.320)

binds here, Irix5b, can no longer bind because it's an AT rich interacting domain.

Lex Fridman (1:44:42.720)

And as soon as I change the T into a C, it can no longer bind because it's no longer

Manolis Kellis (1:44:46.440)

AT rich.

Lex Fridman (1:44:47.440)

But doesn't that mean that you're able to use the energy more efficiently?

Lex Fridman (1:44:52.280)

You're not generating heat or is that?

Lex Fridman (1:44:54.120)

That means I can eat less and get around just fine.

Manolis Kellis (1:44:56.920)

Yes.

Lex Fridman (1:44:57.920)

Yeah.

Lex Fridman (1:44:58.920)

So that's a feature actually.

Lex Fridman (1:44:59.920)

It's a feature in a food scarce environment.

Manolis Kellis (1:45:02.040)

Yeah.

Lex Fridman (1:45:03.040)

But if we're all starving, I'm doing great.

Manolis Kellis (1:45:05.160)

If we all have access to massive amounts of food, I'm obese basically.

Manolis Kellis (1:45:09.360)

That's taken us to the entire process of then understanding that why mitochondria and then

Manolis Kellis (1:45:14.920)

the lipids are both, even though distant, are somehow involved.

Lex Fridman (1:45:18.600)

Different sides of the same coin.

Lex Fridman (1:45:20.760)

And you basically choose to store energy or you can choose to burn energy.

Lex Fridman (1:45:24.000)

And then all of that is involved in the puzzle of obesity.

Lex Fridman (1:45:27.800)

And that's what's fascinating, right?

Manolis Kellis (1:45:29.760)

Here we are in 2007, discovering the strongest genetic association with obesity and knowing

Manolis Kellis (1:45:35.360)

nothing about how it works for almost 10 years.

Manolis Kellis (1:45:39.460)

For 10 years, everybody focused on this FTO gene and they were like, oh, it must have

Manolis Kellis (1:45:43.840)

to do something with RNA modification.

Lex Fridman (1:45:46.240)

And it's like, no, it has nothing to do with the function of FTO.

Manolis Kellis (1:45:50.760)

It has everything to do with all of these other processes.

Lex Fridman (1:45:53.880)

And suddenly the moment you solve that puzzle, which is a multiyear effort by the way, a

Manolis Kellis (1:45:58.680)

tremendous effort by Melina and many, many others.

Lex Fridman (1:46:01.880)

So this tremendous effort basically led us to recognize this circuitry.

Manolis Kellis (1:46:07.160)

You went from having some 89 common variants associated in that region of the DNA sitting

Lex Fridman (1:46:12.500)

on top of this gene to knowing the whole circuitry.

Manolis Kellis (1:46:17.840)

When you know the circuitry, you can now go crazy.

Lex Fridman (1:46:21.160)

You can now start intervening at every level.

Manolis Kellis (1:46:24.480)

You can start intervening at the arid 5B level.

Lex Fridman (1:46:27.240)

You can start intervening with CRISPR Cas9 at the single SNP level.

Manolis Kellis (1:46:31.280)

You can start intervening at iRx3 and iRx5 directly there.

Manolis Kellis (1:46:34.860)

You can start intervening at the thermogenesis level because you know the pathway.

Manolis Kellis (1:46:38.400)

You can start intervening at the differentiation level where the decision to make either white

Manolis Kellis (1:46:45.280)

fat or beige fat, the energy burning beige fat is made developmentally in the first three

Manolis Kellis (1:46:51.500)

days of differentiation of your adipocytes.

Lex Fridman (1:46:54.040)

So as they're differentiating, you basically can choose to make fat burning machines or

Manolis Kellis (1:46:57.720)

fat storing machines.

Lex Fridman (1:46:59.320)

And sort of that's how you populate your fat.

Manolis Kellis (1:47:02.320)

You basically can now go in pharmaceutical and do all of that.

Lex Fridman (1:47:05.880)

And in our paper, we actually did all of that.

Manolis Kellis (1:47:09.400)

We went in and manipulated every single aspect.

Manolis Kellis (1:47:12.320)

At the nucleotide level, we use CRISPR Cas9 genome editing to basically take primary adipocytes

Manolis Kellis (1:47:18.200)

from risk and non risk individuals and show that by editing that one nucleotide out of

Manolis Kellis (1:47:24.080)

3.2 billion nucleotides in the human genome, you could then flip between an obese phenotype

Lex Fridman (1:47:29.600)

and a lean phenotype like a switch.

Manolis Kellis (1:47:31.500)

You can basically take my cells that are non thermogenizing and just flip into thermogenizing

Manolis Kellis (1:47:36.240)

cells by changing one nucleotide.

Lex Fridman (1:47:38.640)

It's mind boggling.

Manolis Kellis (1:47:40.080)

It's so inspiring that this puzzle could be solved in this way and it feels within reach

Lex Fridman (1:47:44.880)

to then be able to crack the problem of some of these diseases.

Lex Fridman (1:47:50.560)

What are the technologies, the tools that came along that made this possible?

Lex Fridman (1:48:00.480)

What are you excited about?

Lex Fridman (1:48:01.980)

Maybe if we just look at the buffet of things that you've kind of mentioned, what's involved?

Lex Fridman (1:48:08.080)

What should we be excited about?

Lex Fridman (1:48:09.520)

What are you excited about?

Lex Fridman (1:48:11.460)

I love that question because there's so much ahead of us.

Manolis Kellis (1:48:14.040)

There's so, so much.

Lex Fridman (1:48:18.600)

So basically solving that one locus required massive amounts of knowledge that we have

Manolis Kellis (1:48:24.000)

been building across the years through the epigenome, through the comparative genomics

Manolis Kellis (1:48:28.220)

to find out the causal variant and the controller regulatory motif through the conserved circuitry.

Manolis Kellis (1:48:35.400)

It required knowing these regulatory genomic wiring.

Manolis Kellis (1:48:38.580)

It required high C of these sort of topologically associated domains to basically find these

Manolis Kellis (1:48:42.980)

long range interaction.

Manolis Kellis (1:48:44.600)

It required EQTLs of these sort of genetic perturbation of these intermediate gene phenotypes.

Manolis Kellis (1:48:51.160)

It required all of the arsenal of tools that I've been describing was put together for

Lex Fridman (1:48:55.640)

one locus.

Lex Fridman (1:48:57.240)

And this was a massive team effort, huge investment in time, energy, money, effort, intellectual,

Lex Fridman (1:49:05.840)

everything.

Manolis Kellis (1:49:06.840)

You're referring to, I'm sorry, just for the obesity one.

Lex Fridman (1:49:09.640)

Yeah, this one paper.

Manolis Kellis (1:49:10.640)

This one single paper.

Lex Fridman (1:49:11.640)

This one single locus.

Manolis Kellis (1:49:12.640)

I would like to say that this is a paper about one nucleotide in the human genome, about

Lex Fridman (1:49:16.640)

one bit of information, C versus T in the human genome.

Manolis Kellis (1:49:20.560)

That's one bit of information and we have 3.2 billion nucleotides to go through.

Lex Fridman (1:49:25.320)

So how do you do that systematically?

Manolis Kellis (1:49:29.240)

I am so excited about the next phase of research because the technologies that my group and

Manolis Kellis (1:49:35.000)

many other groups have developed allows us to now do this systematically, not just one

Manolis Kellis (1:49:40.080)

locus at a time, but thousands of loci at a time.

Lex Fridman (1:49:45.120)

So let me describe some of these technologies.

Manolis Kellis (1:49:48.000)

The first one is automation and robotics.

Lex Fridman (1:49:52.420)

So basically, you know, we talked about how you can take all of these molecules and see

Lex Fridman (1:49:58.240)

which of these molecules are targeting each of these genes and what do they do?

Lex Fridman (1:50:02.200)

So you can basically now screen through millions of molecules through thousands and thousands

Lex Fridman (1:50:07.700)

and thousands of plates, each of which has thousands and thousands and thousands of molecules,

Manolis Kellis (1:50:12.880)

every single time testing, you know, all of these genes and asking which of these molecules

Manolis Kellis (1:50:20.560)

perturb these genes.

Lex Fridman (1:50:22.000)

So that's technology number one, automation and robotics.

Manolis Kellis (1:50:25.880)

Technology number two is parallel readouts.

Lex Fridman (1:50:29.280)

So instead of perturbing one locus and then asking if I use CRISPR Cas9 on this enhancer

Manolis Kellis (1:50:35.880)

to basically use dCas9 to turn on or turn off the enhancer, or if I use CRISPR Cas9

Lex Fridman (1:50:41.280)

on the SNP to basically change that one SNP at a time, then what happens?

Lex Fridman (1:50:46.620)

But we have 120,000 disease associated SNPs that we want to test.

Lex Fridman (1:50:52.760)

We don't want to spend 120,000 years doing it.

Lex Fridman (1:50:57.220)

So what do we do?

Manolis Kellis (1:50:58.920)

We've basically developed this technology for massively parallel reporter assays, MPRA.

Lex Fridman (1:51:07.240)

So in collaboration with Tarsha Mikkelsen, Eric Lander, I mean, Jason Durie's group has

Lex Fridman (1:51:11.240)

done a lot of that.

Lex Fridman (1:51:12.240)

So there's a lot of groups that basically have developed technologies for testing 10,000

Lex Fridman (1:51:19.380)

genetic variants at a time.

Lex Fridman (1:51:21.420)

How do you do that?

Manolis Kellis (1:51:23.000)

You know, we talked about microarray technology, the ability to synthesize these huge microarrays

Manolis Kellis (1:51:28.880)

that allow you to do all kinds of things like measure gene expression by hybridization,

Manolis Kellis (1:51:33.880)

by measuring the genotype of a person, by looking at hybridization with one version

Manolis Kellis (1:51:38.100)

with a T versus the other version with a C, and then sort of figuring out that I am a

Manolis Kellis (1:51:43.400)

risk carrier for obesity based on these differential hybridization in my genome that says, oh,

Manolis Kellis (1:51:49.820)

you seem to only have this allele or you seem to have that allele.

Lex Fridman (1:51:53.320)

These can also be used to systematically synthesize small fragments of DNA.

Lex Fridman (1:51:59.400)

So you can basically synthesize these 150 nucleotide long fragments across 450,000 spots

Lex Fridman (1:52:07.800)

at a time.

Manolis Kellis (1:52:10.240)

You can now take the result of that synthesis, which basically works through all of these

Lex Fridman (1:52:15.820)

sort of layers of adding one nucleotide at a time.

Manolis Kellis (1:52:18.760)

You can basically just type it into your computer and order it, and you can basically order

Lex Fridman (1:52:24.000)

10,000 or 100,000 of these small DNA segments at a time.

Lex Fridman (1:52:30.740)

And that's where awesome molecular biology comes in.

Manolis Kellis (1:52:33.360)

You can basically take all these segments, have a common start and end barcode or sort

Manolis Kellis (1:52:38.840)

of like Gator, just like pieces of a puzzle.

Lex Fridman (1:52:42.120)

You can make the same end piece and the same start piece for all of them.

Lex Fridman (1:52:48.000)

And you can now use plasmids, which are these extra chromosomal small DNA circular segments

Lex Fridman (1:52:57.960)

that are basically inhabiting all our, all our genomes.

Manolis Kellis (1:53:00.560)

We basically have, you know, plasmids from floating around and bacteria use plasmids

Lex Fridman (1:53:05.200)

for transferring DNA.

Lex Fridman (1:53:07.060)

And that's where they put a lot of antibiotic resistance genes.

Lex Fridman (1:53:10.720)

So they can easily transfer them from one bacterium to the other.

Manolis Kellis (1:53:14.200)

After one bacterium evolves a gene to be resistant to a particular antibiotic, it basically says

Lex Fridman (1:53:20.280)

to all its friends, Hey, here's that sort of DNA piece.

Manolis Kellis (1:53:24.760)

We can now coopt these plasmids into human cells.

Manolis Kellis (1:53:28.440)

You can basically make a human cell culture and add plasmids to that human cell culture

Manolis Kellis (1:53:34.000)

that contain the things that you want to test.

Lex Fridman (1:53:38.120)

You now have this library of 450,000 elements.

Manolis Kellis (1:53:41.320)

You can insert them each into the common plasmid and then test them in millions of cells in

Lex Fridman (1:53:47.880)

parallel.

Lex Fridman (1:53:48.880)

And the common plasmid is all the same before you add it.

Lex Fridman (1:53:51.160)

Exactly.

Manolis Kellis (1:53:52.160)

The rest of the plasmid is the same.

Lex Fridman (1:53:53.300)

So it's, it's called an epizomal reporter assay.

Manolis Kellis (1:53:57.640)

Epizome means not inside the genome.

Lex Fridman (1:53:59.720)

It's sort of outside the chromosomes.

Lex Fridman (1:54:01.560)

So it's an epizomal assay that allows you to have a variable region where you basically

Manolis Kellis (1:54:06.200)

test 10,000 different enhancers and you have a common region which basically has the same

Manolis Kellis (1:54:11.720)

reporter gene.

Lex Fridman (1:54:13.720)

You now can do some very cool molecular biology.

Manolis Kellis (1:54:16.600)

You can basically take the 450,000 elements that you've generated and you have a piece

Lex Fridman (1:54:21.960)

of the puzzle here, piece of the puzzle here, which is identical.

Lex Fridman (1:54:24.440)

So they're compatible with that plasmid.

Manolis Kellis (1:54:27.060)

You can chop them up in the middle to separate a barcode reporter from the enhancer and in

Manolis Kellis (1:54:32.840)

the middle put the same gene again using the same piece of the puzzle.

Manolis Kellis (1:54:36.920)

You now can have a barcode readout of what is the impact of 10,000 different versions

Manolis Kellis (1:54:42.960)

of an enhancer on gene expression.

Lex Fridman (1:54:46.600)

So we're not doing one experiment, we're doing 10,000 experiments.

Lex Fridman (1:54:50.680)

And those 10,000 can be 5,000 of different loci and each of them in two versions, risk

Lex Fridman (1:54:58.580)

or non risk.

Manolis Kellis (1:55:00.260)

I can now test tens of thousands.

Lex Fridman (1:55:01.920)

Just a little hypothesis.

Manolis Kellis (1:55:02.920)

Exactly.

Lex Fridman (1:55:03.920)

And then you can do 10,000 and we can test 10,000 hypothesis at once.

Lex Fridman (1:55:08.880)

How hard is it to generate those 10,000?

Lex Fridman (1:55:11.360)

Trivial.

Manolis Kellis (1:55:12.360)

Trivial.

Lex Fridman (1:55:13.360)

But it's biology.

Manolis Kellis (1:55:14.360)

No, no.

Lex Fridman (1:55:15.360)

Generating the 10,000 is trivial because you basically add, it's biotechnology.

Manolis Kellis (1:55:20.740)

You basically have these arrays that add one nucleotide at a time at every spot.

Lex Fridman (1:55:26.560)

So it's printing and so you're able to, you're able to control.

Manolis Kellis (1:55:30.680)

Yeah.

Lex Fridman (1:55:31.680)

Is it super costly?

Lex Fridman (1:55:32.800)

Is it?

Lex Fridman (1:55:33.800)

10,000 bucks.

Lex Fridman (1:55:34.800)

So this isn't millions.

Lex Fridman (1:55:35.800)

10,000 bucks for 10,000 experiments sounds like the right, you know.

Manolis Kellis (1:55:39.200)

I mean, so that's super, that's exciting because you don't have to do one thing at a time.

Manolis Kellis (1:55:44.100)

You can now use that technology, these massively parallel reporter assays to test 10,000 locations

Manolis Kellis (1:55:49.280)

at a time.

Lex Fridman (1:55:51.440)

We've made multiple modifications to that technology.

Manolis Kellis (1:55:55.160)

One was sharper MPRA, which stands for, you know, basically getting a higher resolution

Manolis Kellis (1:56:04.080)

view by tiling these, these elements so you can see where along the region of control

Manolis Kellis (1:56:14.800)

are they acting.

Lex Fridman (1:56:16.140)

And we made another modification called Hydra for high, you know, definition regulatory

Manolis Kellis (1:56:23.240)

annotation or something like that, which basically allows you to test 7 million of these at a

Lex Fridman (1:56:30.080)

time by sort of cutting them directly from the DNA.

Lex Fridman (1:56:32.960)

So instead of synthesizing, which basically has the limit of 450,000 that you can synthesize

Manolis Kellis (1:56:37.420)

at a time, we basically said, Hey, if we want to test all accessible regions of the genome,

Manolis Kellis (1:56:42.600)

let's just do an experiment that cuts accessible regions.

Manolis Kellis (1:56:45.620)

Let's take those accessible regions, put them all with the same end joints of the puzzles,

Lex Fridman (1:56:51.520)

and then now use those to create a much, much larger array of things that you can test.

Lex Fridman (1:56:59.680)

And then tiling all of these regions, you can then pinpoint what are the driver nucleotides,

Lex Fridman (1:57:04.160)

what are the elements, how are they acting across 7 million experiments at a time.

Lex Fridman (1:57:07.520)

So basically this is all the same family of technology where you're basically using these

Manolis Kellis (1:57:12.580)

parallel readouts of the barcodes.

Lex Fridman (1:57:15.900)

And then to do this, we used a technology called StarSeq for self transcribing reporter

Manolis Kellis (1:57:23.240)

assays, a technology developed by Alex Stark, my former postdoc, who's now API over in Vienna.

Lex Fridman (1:57:30.140)

So we basically coupled the StarSeq, the self transcribing reporters where the enhancer

Manolis Kellis (1:57:37.240)

can be part of the gene itself.

Lex Fridman (1:57:39.040)

So instead of having a separate barcode, that enhancer basically acts to turn on the gene

Lex Fridman (1:57:43.600)

and it's transcribed as part of the gene.

Lex Fridman (1:57:46.080)

So you don't have to have the two separate parts.

Manolis Kellis (1:57:47.640)

Exactly.

Lex Fridman (1:57:48.640)

So you can just read them directly.

Lex Fridman (1:57:49.640)

So there's a constant improvements in this whole process.

Lex Fridman (1:57:52.680)

By the way, generating all these options, is it basically brute force?

Lex Fridman (1:57:57.160)

How much human intuition is?

Manolis Kellis (1:57:58.680)

Oh gosh, of course it's human intuition and human creativity and incorporating all of

Manolis Kellis (1:58:04.040)

the input data sets.

Lex Fridman (1:58:06.040)

Because again, the genome is enormous.

Manolis Kellis (1:58:08.440)

3.2 billion, you don't want to test that.

Lex Fridman (1:58:11.040)

You basically use all of these tools that I've talked about already.

Manolis Kellis (1:58:14.280)

You generate your top favorite 10,000 hypothesis, and then you go and test all 10,000.

Lex Fridman (1:58:19.920)

And then from what comes out, you can then go to the next step.

Lex Fridman (1:58:24.080)

So that's technology number two.

Lex Fridman (1:58:25.920)

So technology number one is robotics, automation, where you have thousands of wells and you

Manolis Kellis (1:58:30.440)

constantly test them.

Manolis Kellis (1:58:32.140)

The second technology is instead of having wells, you have these massively parallel readouts

Manolis Kellis (1:58:37.320)

in sort of these pooled assays.

Manolis Kellis (1:58:40.000)

The third technology is coupling CRISPR perturbations with these single cell RNA readouts.

Lex Fridman (1:58:51.260)

So let me make another parenthesis here to describe now single cell RNA sequencing.

Lex Fridman (1:58:57.880)

So what does single cell RNA sequencing mean?

Lex Fridman (1:58:59.720)

So RNA sequencing is what has been traditionally used, well, traditionally the last 20 years,

Lex Fridman (1:59:07.760)

ever since the advent of next generation sequencing.

Lex Fridman (1:59:10.200)

So basically before RNA expression profiling was based on these microarrays.

Lex Fridman (1:59:14.620)

The next technology after that was based on sequencing.

Lex Fridman (1:59:17.500)

So you chop up your RNA and you just sequence small molecules, just like you would sequence

Manolis Kellis (1:59:22.840)

a genome, basically reverse transcribe the small RNAs into DNA, and you sequence that

Manolis Kellis (1:59:28.040)

DNA in order to get the number of sequencing reads corresponding to the expression level

Lex Fridman (1:59:35.600)

of every gene in the genome.

Manolis Kellis (1:59:37.480)

You now have RNA sequencing.

Lex Fridman (1:59:39.680)

How do you go to single cell RNA sequencing?

Manolis Kellis (1:59:42.520)

That technology also went through stages of evolution.

Lex Fridman (1:59:45.880)

The first was microfluidics.

Manolis Kellis (1:59:48.120)

You basically had these, or even chambers, you basically had these ways of isolating

Lex Fridman (1:59:52.940)

individual cells, putting them into a well for every one of these cells.

Lex Fridman (1:59:57.320)

So you have 384 well plates and you now do 384 parallel reactions to measure the expression

Lex Fridman (20:02.420)

But the moment you sort of break up that very long path into smaller levels, you can basically

Manolis Kellis (20:07.400)

say from genetics, what are the epigenomic alterations at the level of gene regulatory

Lex Fridman (20:13.480)

elements where that genetic variant perturbs the control region nearby.

Manolis Kellis (20:19.160)

That effect is much larger.

Lex Fridman (20:21.840)

You mean much larger in terms of this down the line impact or?

Manolis Kellis (20:25.480)

It's much larger in terms of the measurable effect, this A versus B variance is actually

Lex Fridman (20:31.120)

so much cleanly defined when you go to the shorter branches.

Manolis Kellis (20:35.800)

Because for one genetic variant to affect Alzheimer's, that's a very long path.

Manolis Kellis (20:40.760)

That basically means that in the context of millions of these 6 million variants that

Manolis Kellis (20:43.940)

every one of us carries, that one single nucleotide has a detectable effect all the way to the

Lex Fridman (20:51.040)

end.

Manolis Kellis (20:52.040)

I mean, it's just mind boggling that that's even possible, but indeed there are such effects.

Lex Fridman (20:57.700)

So the hope is, or the most scientifically speaking, the most effective place where to

Manolis Kellis (21:03.000)

detect the alteration that results in disease is earlier on in the pipeline, as early as

Lex Fridman (21:10.640)

possible.

Manolis Kellis (21:11.640)

It's a trade off.

Manolis Kellis (21:12.680)

If you go very early on in the pipeline, now each of these epigenomic alterations, for

Manolis Kellis (21:17.800)

example, this enhancer control region is active maybe 50% less, which is a dramatic effect.

Manolis Kellis (21:25.500)

Now you can ask, well, how much does changing one regulatory region in the genome in one

Lex Fridman (21:29.680)

cell type change disease?

Lex Fridman (21:31.280)

Well, that path is now long.

Lex Fridman (21:33.920)

So if you instead look at expression, the path between genetic variation and the expression

Manolis Kellis (21:39.680)

of one gene goes through many enhancer regions, and therefore it's a subtler effect at the

Manolis Kellis (21:44.560)

gene level.

Lex Fridman (21:45.560)

But then now you're closer because one gene is acting in the context of only 20,000 other

Manolis Kellis (21:51.360)

genes as opposed to one enhancer acting in the context of 2 million other enhancers.

Lex Fridman (21:57.200)

So you basically now have genetic, epigenomic, the circuitry, transcriptomic, the gene expression

Manolis Kellis (22:04.040)

control, and then cellular, where you can basically say, I can measure various properties

Lex Fridman (22:09.600)

of those cells.

Lex Fridman (22:11.160)

What is the calcium influx rate when I have this genetic variation?

Lex Fridman (22:17.560)

What is the synaptic density?

Lex Fridman (22:19.760)

What is the electric impulse conductivity and so on and so forth?

Lex Fridman (22:24.500)

So you can measure things along this path to disease, and you can also measure endophenotypes.

Manolis Kellis (22:32.660)

You can basically measure your brain activity.

Lex Fridman (22:37.460)

You can do imaging in the brain.

Manolis Kellis (22:39.680)

You can basically measure, I don't know, the heart rate, the pulse, the lipids, the amount

Lex Fridman (22:44.440)

of blood secreted and so on and so forth.

Lex Fridman (22:46.700)

And then through all of that, you can basically get at the path to causality, the path to

Lex Fridman (22:52.760)

disease.

Lex Fridman (22:55.320)

And is there something beyond cellular?

Lex Fridman (22:57.680)

So you mentioned lifestyle interventions or changes as a way to, or like be able to prescribe

Manolis Kellis (23:05.480)

changes in lifestyle.

Lex Fridman (23:07.840)

Like what about organs?

Lex Fridman (23:09.360)

What about like the function of the body as a whole?

Lex Fridman (23:13.200)

Yeah, absolutely.

Lex Fridman (23:14.200)

So basically when you go to your doctor, they always measure, you know, your pulse.

Lex Fridman (23:18.200)

They always measure your height.

Manolis Kellis (23:19.200)

They always measure your weight, you know, your BMI.

Lex Fridman (23:21.240)

So basically these are just very basic variables.

Lex Fridman (23:24.180)

But with digital devices nowadays, you can start measuring hundreds of variables for

Lex Fridman (23:27.960)

every individual.

Manolis Kellis (23:29.600)

You can basically also phenotype cognitively through tests, Alzheimer's patients.

Manolis Kellis (23:37.300)

There are cognitive tests that you can measure, that you typically do for cognitive decline,

Manolis Kellis (23:43.720)

these mini mental observations that you have specific questions to.

Lex Fridman (23:48.500)

You can think of sort of enlarging the set of cognitive tests.

Lex Fridman (23:51.980)

So in the mouse, for example, you do experiments for how do they get out of mazes?

Lex Fridman (23:55.760)

How do they find food?

Manolis Kellis (23:57.280)

Whether they recall a fear, whether they shake in a new environment and so on and so forth.

Manolis Kellis (24:02.440)

In the human, you can have much, much richer phenotypes where you can basically say not

Manolis Kellis (24:06.560)

just imaging at the organ level and all kinds of other activities at the organ level, but

Lex Fridman (24:13.920)

you can also do at the organism level, you can do behavioral tests.

Lex Fridman (24:19.480)

And how did they do on empathy?

Lex Fridman (24:21.120)

How did they do on memory?

Lex Fridman (24:22.920)

How did they do on longterm memory versus short term memory?

Lex Fridman (24:26.160)

And so on and so forth.

Manolis Kellis (24:27.160)

I love how you're calling that phenotype.

Lex Fridman (24:28.760)

I guess it is.

Manolis Kellis (24:29.760)

It is.

Lex Fridman (24:31.040)

But like your behavior patterns that might change over a period of a life, your ability

Manolis Kellis (24:37.880)

to remember things, your ability to be empathetic or emotionally, your intelligence perhaps

Lex Fridman (24:44.560)

even.

Manolis Kellis (24:45.560)

Yeah, but intelligence has hundreds of variables.

Manolis Kellis (24:47.160)

You can be your math intelligence, your literary intelligence, your puzzle solving intelligence,

Manolis Kellis (24:50.720)

your logic.

Lex Fridman (24:51.720)

It could be like hundreds of things.

Lex Fridman (24:52.880)

And all of that, we're able to measure that better and better and all that could be connected

Lex Fridman (24:57.440)

to the entire pipeline somehow.

Manolis Kellis (24:58.920)

We used to think of each of these as a single variable like intelligence.

Lex Fridman (25:01.840)

I mean, that's ridiculous.

Manolis Kellis (25:03.380)

It's basically dozens of different genes that are controlling every single variable.

Manolis Kellis (25:10.880)

You can basically think of, imagine us in a video game where every one of us has measures

Manolis Kellis (25:16.040)

of strength, stamina, energy left and so on and so forth.

Lex Fridman (25:20.960)

But you could click on each of those five bars that are just the main bars and each

Manolis Kellis (25:24.440)

of those will just give you then hundreds of bars and can basically say, okay, great

Manolis Kellis (25:28.560)

for my machine learning task, I want someone who, a human who has these particular forms

Manolis Kellis (25:36.200)

of intelligence.

Lex Fridman (25:37.200)

I require now these 20 different things.

Lex Fridman (25:40.620)

And then you can combine those things and then relate them to of course performance

Manolis Kellis (25:45.000)

in a particular task, but you can also relate them to genetic variation that might be affecting

Manolis Kellis (25:50.820)

different parts of the brain.

Manolis Kellis (25:52.800)

For example, your frontal cortex versus your temporal cortex versus your visual cortex

Lex Fridman (25:56.600)

and so on and so forth.

Lex Fridman (25:58.040)

So genetic variation that affects expression of genes in different parts of your brain

Manolis Kellis (26:02.520)

can basically affect your music ability, your auditory ability, your smell, just dozens

Manolis Kellis (26:08.920)

of different phenotypes can be broken down into hundreds of cognitive variables and then

Manolis Kellis (26:15.980)

relate each of those to thousands of genes that are associated with them.

Lex Fridman (26:20.520)

So somebody who loves RPGs or playing games, there's too few variables that we can control.

Lex Fridman (26:28.440)

So I'm excited if we're in fact living in a simulation and this is a video game, I'm

Lex Fridman (26:32.680)

excited by the quality of the video game.

Manolis Kellis (26:37.240)

The game designer did a hell of a good job.

Lex Fridman (26:39.760)

So we're impressed.

Manolis Kellis (26:40.760)

Oh, I don't know.

Lex Fridman (26:41.760)

The sunset last night was a little unrealistic.

Manolis Kellis (26:43.800)

Yeah.

Lex Fridman (26:44.800)

Yeah.

Manolis Kellis (26:45.800)

The graphics.

Lex Fridman (26:46.800)

Exactly.

Manolis Kellis (26:47.800)

Come on, NVIDIA.

Manolis Kellis (26:48.800)

To zoom back out, we've been talking about the genetic origins of diseases, but I think

Manolis Kellis (26:54.480)

it's fascinating to talk about what are the most important diseases to understand and

Lex Fridman (27:01.080)

especially as it connects to the things that you're working on.

Lex Fridman (27:05.320)

So it's very difficult to think about important diseases to understand.

Lex Fridman (27:08.840)

There's many metrics of importance.

Manolis Kellis (27:10.500)

One is lifestyle impact.

Lex Fridman (27:12.360)

I mean, if you look at COVID, the impact on lifestyle has been enormous.

Lex Fridman (27:16.440)

So understanding COVID is important because it has impacted the wellbeing in terms of

Manolis Kellis (27:23.080)

ability to have a job, ability to have an apartment, ability to go to work, ability

Manolis Kellis (27:27.280)

to have a mental circle of support and all of that for millions of Americans, like huge,

Lex Fridman (27:34.480)

huge impact.

Lex Fridman (27:35.520)

So that's one aspect of importance.

Lex Fridman (27:37.000)

So basically mental disorders, Alzheimer's has a huge importance in the wellbeing of

Manolis Kellis (27:42.480)

Americans.

Lex Fridman (27:44.040)

Whether or not it kills someone for many, many years, it has a huge impact.

Lex Fridman (27:48.220)

So the first measure of importance is just wellbeing.

Lex Fridman (27:52.360)

Impact on the quality of life.

Manolis Kellis (27:53.780)

Impact on the quality of life, absolutely.

Lex Fridman (27:55.860)

The second metric, which is much easier to quantify is deaths.

Lex Fridman (28:00.160)

What is the number one killer?

Lex Fridman (28:01.920)

The number one killer is actually heart disease.

Manolis Kellis (28:04.760)

It is actually killing 650,000 Americans per year.

Lex Fridman (28:10.700)

Number two is cancer with 600,000 Americans.

Manolis Kellis (28:14.280)

Number three, far, far down the list is accidents, every single accident combined.

Lex Fridman (28:19.600)

So basically you read the news, accidents, like there was a huge car crash all over the

Manolis Kellis (28:24.720)

news.

Lex Fridman (28:25.800)

But the number of deaths, number three by far, 167,000.

Manolis Kellis (28:31.320)

Core respiratory disease.

Lex Fridman (28:32.800)

So that's asthma, not being able to breathe and so on and so forth, 160,000 Alzheimer's

Manolis Kellis (28:39.160)

number five with 120,000 and then stroke, brain aneurysms and so on and so forth, that's

Lex Fridman (28:45.040)

147,000 diabetes and metabolic disorders, et cetera.

Manolis Kellis (28:49.720)

That's 85,000.

Manolis Kellis (28:51.140)

The flu is 60,000, suicide, 50,000 and then overdose, et cetera, you know, goes further

Manolis Kellis (28:58.960)

down the list.

Lex Fridman (29:00.120)

So of course COVID has creeped up to be the number three killer this year with, you know,

Manolis Kellis (29:06.620)

more than 100,000 Americans and counting.

Lex Fridman (29:11.360)

And you know, but if you think about sort of what do we use, what are the most important

Manolis Kellis (29:16.560)

diseases, you have to understand both the quality of life and the sheer number of deaths

Lex Fridman (29:22.720)

and just numbers of years lost if you wish.

Lex Fridman (29:25.560)

And each of these diseases you can think of as, and also including terrorist attacks and

Manolis Kellis (29:30.960)

school shootings, for example, things which lead to fatalities, you can look at as problems

Manolis Kellis (29:39.200)

that could be solved.

Lex Fridman (29:41.480)

And some problems are harder to solve than others.

Manolis Kellis (29:44.080)

I mean, that's part of the equation.

Lex Fridman (29:46.860)

So maybe if you look at these diseases, if you look at heart disease or cancer or Alzheimer's

Manolis Kellis (29:52.960)

or just like schizophrenia and obesity, Debbie, like not necessarily things that kill you,

Lex Fridman (29:59.800)

but affect the quality of life, which problems are solvable, which aren't, which are harder

Manolis Kellis (2:00:03.280)

of 384 cells.

Lex Fridman (2:00:05.660)

That sounds amazing and it was amazing, but we want to do a million cells.

Lex Fridman (2:00:11.320)

How do you go from these wells to a million cells?

Lex Fridman (2:00:14.120)

You can't.

Lex Fridman (2:00:15.640)

So what the next technology was after that is instead of using a well for every reaction,

Lex Fridman (2:00:21.660)

you now use a lipid droplet for every reaction.

Lex Fridman (2:00:26.280)

So you use micro droplets as reaction chambers to basically amplify RNA.

Lex Fridman (2:00:33.660)

So here's the idea.

Manolis Kellis (2:00:34.660)

You basically have microfluidics where you basically have every single cell coming down

Manolis Kellis (2:00:39.280)

one tube in your microfluidics and you have little bubbles getting created in the other

Manolis Kellis (2:00:44.040)

way with specific primers that mark every cell with its own barcode.

Manolis Kellis (2:00:49.360)

You basically couple the two and you end up with little bubbles that have a cell and tons

Manolis Kellis (2:00:55.040)

of markers for that cell.

Manolis Kellis (2:00:57.400)

You now mark up all of the RNA for that one cell with the same exact barcode and you then

Manolis Kellis (2:01:03.880)

lyse all of the droplets and you sequence the heck out of that and you have for every

Lex Fridman (2:01:09.360)

RNA molecule, a unique identifier that tells you what cell was it on.

Manolis Kellis (2:01:12.880)

That is such good engineering, microfluidics and using some kind of primer to put a label

Lex Fridman (2:01:20.840)

on the thing.

Manolis Kellis (2:01:21.840)

I mean, you're making it sound easy.

Lex Fridman (2:01:24.080)

I assume it's beautiful, but it's gorgeous.

Lex Fridman (2:01:27.400)

So there's the next generation.

Lex Fridman (2:01:29.560)

So that's the second generation.

Manolis Kellis (2:01:31.120)

Next generation is forget the microfluidics altogether.

Lex Fridman (2:01:34.000)

Just use big bottles.

Lex Fridman (2:01:35.000)

How can you possibly do that with big bottles?

Lex Fridman (2:01:37.960)

So here's the idea.

Manolis Kellis (2:01:39.400)

You dissociate all of your cells or all of your nuclei from complex cells like brain

Lex Fridman (2:01:43.680)

cells that are very long and sticky so you can't do that.

Manolis Kellis (2:01:48.240)

If you have blood cells or if you have neuronal nuclei or brain nuclei, you can basically

Lex Fridman (2:01:52.520)

dissociate let's say a million cells.

Manolis Kellis (2:01:56.160)

You now want to add a unique barcode, a unique barcode in each one of a million cells using

Lex Fridman (2:02:01.720)

only big bottles.

Lex Fridman (2:02:02.720)

How can you possibly do that?

Lex Fridman (2:02:04.440)

Sounds crazy, but here's the idea.

Manolis Kellis (2:02:07.320)

You use a hundred of these bottles, you randomly shuffle all your million cells and you throw

Lex Fridman (2:02:13.880)

them into those hundred bottles randomly, completely randomly.

Manolis Kellis (2:02:17.180)

You add one barcode out of a hundred to every one of the cells.

Lex Fridman (2:02:21.560)

You then you now take them all out.

Manolis Kellis (2:02:23.560)

You shuffle them again and you throw them again into the same hundred bottles.

Lex Fridman (2:02:28.440)

But now in a different randomization and you add a second barcode.

Lex Fridman (2:02:33.960)

So every cell now has two barcodes.

Lex Fridman (2:02:36.880)

You take them out again, you shuffle them and you throw them back in.

Manolis Kellis (2:02:40.280)

Another third barcode is adding randomly from the same hundred barcodes.

Manolis Kellis (2:02:47.480)

You've now labeled every cell probabilistically based on the unique path that he took of which

Manolis Kellis (2:02:53.920)

of a hundred bottles did he go for the first time, which of a hundred bottles the second

Lex Fridman (2:02:56.880)

time and which of a hundred bottles the third time.

Manolis Kellis (2:03:00.160)

A hundred times a hundred times a hundred is a million unique barcodes in every single

Lex Fridman (2:03:05.240)

one of these cells without ever using microfluidics.

Manolis Kellis (2:03:09.480)

Very clever.

Lex Fridman (2:03:10.480)

It's beautiful, right?

Manolis Kellis (2:03:11.480)

From a computer science perspective, that's very clever.

Lex Fridman (2:03:12.880)

Yeah.

Lex Fridman (2:03:13.880)

So you now have the single cell sequence technology.

Manolis Kellis (2:03:16.160)

You can use the wells, you can use the bubbles or you can use the bottles and you have way

Manolis Kellis (2:03:22.040)

The bubbles still sound pretty damn cool.

Lex Fridman (2:03:23.680)

The bubbles are awesome.

Lex Fridman (2:03:24.680)

And that's basically the main technology that we're using.

Lex Fridman (2:03:26.640)

So the bubbles is the main technology.

Lex Fridman (2:03:29.680)

So there are kits now that companies just sell to basically carry out single cell RNA

Manolis Kellis (2:03:34.360)

sequencing that you can basically for $2,000, you can basically get 10,000 cells from one

Manolis Kellis (2:03:40.240)

sample.

Lex Fridman (2:03:42.560)

And for every one of those cells, you basically have the transcription of thousands of genes.

Lex Fridman (2:03:49.680)

And you know, of course the data for any one cell is noisy, but being computer scientists,

Manolis Kellis (2:03:54.360)

we can aggregate the data from all of the cells together across thousands of individuals

Manolis Kellis (2:03:58.640)

together to basically make very robust inferences.

Lex Fridman (2:04:02.120)

Okay.

Lex Fridman (2:04:03.120)

So the third technology is basically single cell RNA sequencing that allows you to now

Manolis Kellis (2:04:07.160)

start asking not just what is the brain expression level difference of that genetic variant,

Lex Fridman (2:04:14.400)

but what is the expression difference of that one genetic variant across every single subtype

Lex Fridman (2:04:20.000)

of brain cell?

Lex Fridman (2:04:21.720)

How is the variance changing?

Manolis Kellis (2:04:24.460)

You can't just, you know, with a brain sample, you can just ask about the mean, what is the

Lex Fridman (2:04:29.260)

average expression?

Manolis Kellis (2:04:30.840)

If I instead have 3000 cells that are neurons, I can ask not just what is the neuronal expression.

Manolis Kellis (2:04:38.280)

I can say for layer five excitatory neurons of which I have, I don't know, 300 cells,

Lex Fridman (2:04:44.240)

what is the variance that this genetic variant has?

Lex Fridman (2:04:48.240)

So suddenly it's amazingly more powerful.

Manolis Kellis (2:04:51.000)

I can basically start asking about this middle layer of gene expression at unprecedented

Manolis Kellis (2:04:55.240)

levels.

Lex Fridman (2:04:56.240)

So when you look at the average, it washes out some potentially important signal that

Manolis Kellis (2:05:01.600)

corresponds to ultimately the disease.

Lex Fridman (2:05:04.160)

Completely.

Manolis Kellis (2:05:05.160)

Yeah.

Lex Fridman (2:05:06.160)

So that, I can do that at the RNA level, but I can also do that at the DNA level for the

Manolis Kellis (2:05:10.200)

epigenome.

Lex Fridman (2:05:11.200)

So remember how before I was telling you about all this technology that we're using to probe

Manolis Kellis (2:05:14.760)

the epigenome, one of them is DNA accessibility.

Lex Fridman (2:05:18.160)

So what we're doing in my lab is that from the same dissociation of say a brain sample

Manolis Kellis (2:05:23.200)

where you now have all these tens of thousands of cells floating around, you basically take

Manolis Kellis (2:05:27.480)

half of them to do RNA profiling and the other half to do epigenome profiling, both at the

Manolis Kellis (2:05:32.360)

single cell level.

Lex Fridman (2:05:34.140)

So that allows you to now figure out what are the millions of DNA enhancers that are

Manolis Kellis (2:05:40.340)

accessible in every one of tens of thousands of cells.

Lex Fridman (2:05:45.000)

And computationally, we can now take the RNA and the DNA readouts and group them together

Manolis Kellis (2:05:50.600)

to basically figure out how is every enhancer related to every gene.

Lex Fridman (2:05:57.600)

And remember these sort of enhancer gene linking that we were doing across 833 samples?

Manolis Kellis (2:06:01.720)

833 is awesome, don't get me wrong, but 10 million is way more awesome.

Lex Fridman (2:06:08.240)

So we can now look at correlated activity across 2.3 million enhancers and 20,000 genes

Manolis Kellis (2:06:14.600)

in each of millions of cells to basically start piecing together the regulatory circuitry

Manolis Kellis (2:06:19.860)

of every single type of neuron, every single type of astrocytes, oligodendrocytes, microglial

Manolis Kellis (2:06:25.440)

cell inside the brains of 1,500 individuals that we sample across multiple different brain

Lex Fridman (2:06:32.880)

regions across both DNA and RNA.

Lex Fridman (2:06:36.240)

So that's the data set that my team generated last year alone.

Lex Fridman (2:06:39.600)

So in one year, we basically generated 10 million cells from human brain across a dozen

Manolis Kellis (2:06:46.560)

different disorders, across schizophrenia, Alzheimer's, frontotemporal dementia, Lewy

Manolis Kellis (2:06:51.200)

body dementia, ALS, Huntington's disease, post traumatic stress disorder, autism, bipolar

Manolis Kellis (2:07:01.000)

disorder, healthy aging, et cetera.

Lex Fridman (2:07:04.400)

So it's possible that even just within that data set lie a lot of keys to understanding

Manolis Kellis (2:07:13.120)

these diseases and then be able to like directly leads to then treatment.

Lex Fridman (2:07:18.320)

Correct.

Manolis Kellis (2:07:19.320)

Correct.

Lex Fridman (2:07:20.320)

So basically we are now motivating.

Manolis Kellis (2:07:21.880)

Yeah.

Lex Fridman (2:07:22.880)

So our computational team is in heaven right now and we're looking for people.

Manolis Kellis (2:07:25.680)

I mean, if you have super smart.

Lex Fridman (2:07:29.700)

So this is a very interesting kind of side question.

Lex Fridman (2:07:33.080)

How much of this is biology?

Lex Fridman (2:07:34.680)

How much of this is computation?

Lex Fridman (2:07:36.280)

So you're the head of the computational biology group, but how much of, should you be comfortable

Lex Fridman (2:07:44.080)

with biology to be able to solve some of these problems?

Manolis Kellis (2:07:48.600)

If you just find, if you put several of the hats you were on fundamentally, are you thinking

Lex Fridman (2:07:54.120)

like a computer scientist here?

Manolis Kellis (2:07:56.460)

You have to.

Lex Fridman (2:07:57.460)

This is the only way.

Manolis Kellis (2:07:59.760)

As I said, we are the descendants of the first digital computer.

Lex Fridman (2:08:02.720)

We're trying to understand the digital computer.

Manolis Kellis (2:08:05.000)

We're trying to understand the circuitry, the logic of this digital core computer and

Lex Fridman (2:08:11.240)

all of these analog layers surrounding it.

Lex Fridman (2:08:14.200)

So the case that I've been making is that you cannot think one gene at a time.

Lex Fridman (2:08:19.840)

The traditional biology is dead.

Manolis Kellis (2:08:22.080)

There's no way you cannot solve disease with traditional biology.

Lex Fridman (2:08:24.960)

You need it as a component.

Manolis Kellis (2:08:27.240)

Once you figured out RX3 and RX5, you now can then say, Hey, have you guys worked on

Lex Fridman (2:08:31.840)

those genes with your single gene approach?

Manolis Kellis (2:08:33.880)

We'd love to know everything you know.

Lex Fridman (2:08:35.560)

And if you haven't, we now know how important these genes are.

Manolis Kellis (2:08:38.960)

Let's now launch a single gene program to dissect them and understand them.

Lex Fridman (2:08:43.520)

But you cannot use that as a way to dissect disease.

Manolis Kellis (2:08:46.680)

You have to think genomically.

Manolis Kellis (2:08:48.580)

You have to think from the global perspective and you have to build these circuits systematically.

Lex Fridman (2:08:53.380)

So we need numbers of computer scientists who are interested and willing to dive into

Lex Fridman (2:08:59.220)

these data fully, fully in and extract meaning.

Manolis Kellis (2:09:04.960)

We need computer science people who can understand machine learning and inference and decouple

Lex Fridman (2:09:11.960)

these matrices, come up with super smart ways of dissecting them.

Lex Fridman (2:09:16.360)

But we also need computer scientists who understand biology, who are able to design the next generation

Lex Fridman (2:09:22.880)

of experiments.

Manolis Kellis (2:09:24.660)

Because many of these experiments, no one in their right mind would design them without

Manolis Kellis (2:09:28.760)

thinking of the analytical approach that you would use to deconvolve the data afterwards.

Manolis Kellis (2:09:33.020)

Because it's massive amounts of ridiculously noisy data.

Lex Fridman (2:09:36.640)

And if you don't have the computational pipeline in your head before you even design the experiment,

Manolis Kellis (2:09:42.700)

you would never design the experiment that way.

Lex Fridman (2:09:44.760)

That's brilliant.

Lex Fridman (2:09:45.760)

So in designing the experiment, you have to see the entirety of the computational pipeline.

Lex Fridman (2:09:50.160)

That drives the design.

Manolis Kellis (2:09:52.600)

That even drives the necessity for that design.

Manolis Kellis (2:09:55.560)

Basically, you know, if you didn't have a computer scientist way of thinking, you would

Manolis Kellis (2:10:00.320)

never design these hugely combinatorial, massively parallel experiments.

Lex Fridman (2:10:07.360)

So that's why you need interdisciplinary teams, you need teams.

Lex Fridman (2:10:10.680)

And I want to sort of clarify that what do we mean by computational biology group?

Lex Fridman (2:10:15.200)

The focus is not on computational, the focus is on the biology.

Lex Fridman (2:10:18.880)

So we are a biology group.

Lex Fridman (2:10:20.920)

What type of biology?

Manolis Kellis (2:10:22.680)

Computational biology.

Lex Fridman (2:10:23.680)

That's the type of biology that uses the whole genome.

Manolis Kellis (2:10:27.760)

That's the type of biology that designs experiments, genomic experiments, that can only be interpreted

Lex Fridman (2:10:33.040)

in the context of the whole genome.

Manolis Kellis (2:10:34.600)

Right.

Lex Fridman (2:10:35.600)

So it's philosophically looking at biology as a computer.

Manolis Kellis (2:10:39.800)

Correct.

Lex Fridman (2:10:40.800)

Correct.

Lex Fridman (2:10:41.800)

So which is in the context of the history of biology is a big transformation.

Lex Fridman (2:10:46.280)

Yeah.

Manolis Kellis (2:10:47.280)

Yeah.

Lex Fridman (2:10:48.280)

You can think of the name as what do we do?

Manolis Kellis (2:10:50.200)

Only computation.

Lex Fridman (2:10:51.240)

That's not true.

Lex Fridman (2:10:52.240)

How do we study it?

Lex Fridman (2:10:53.880)

Only computationally.

Manolis Kellis (2:10:54.880)

That is true.

Lex Fridman (2:10:56.520)

So all of these single cell sequencing can now be coupled with the technology that we

Manolis Kellis (2:11:00.480)

talked about earlier for perturbation.

Lex Fridman (2:11:02.920)

So here's the crazy thing.

Manolis Kellis (2:11:04.560)

Instead of using these wells and these robotic systems for doing one drug at a time or for

Manolis Kellis (2:11:10.720)

perturbing one gene at a time in thousands of wells, you can now do this using a pool

Manolis Kellis (2:11:16.880)

of cells and single cell RNA sequencing.

Lex Fridman (2:11:20.120)

How?

Manolis Kellis (2:11:21.120)

You basically can take these perturbations using CRISPR and instead of using a single

Manolis Kellis (2:11:27.960)

guide RNA, you can use a library of guide RNAs generated exactly the same way using

Manolis Kellis (2:11:32.920)

this array technology.

Lex Fridman (2:11:34.480)

So you synthesize a thousand different guide RNAs.

Manolis Kellis (2:11:38.500)

You now take each of these guide RNAs and you insert them in a pool of cells where every

Lex Fridman (2:11:45.720)

cell gets one perturbation.

Lex Fridman (2:11:48.220)

And you use CRISPR editing or CRISPR, so with either CRISPR Cas9 to edit a genome with these

Lex Fridman (2:11:56.720)

thousand perturbations or with the activation or with the repression.

Lex Fridman (2:12:01.400)

And you now can have a single cell readout where every single cell has received one of

Lex Fridman (2:12:07.600)

these modifications.

Lex Fridman (2:12:09.600)

And you can now in massively parallel ways, couple the perturbation and the readout in

Lex Fridman (2:12:17.080)

a single experiment.

Lex Fridman (2:12:18.480)

How are you tracking which perturbations each cell received?

Lex Fridman (2:12:21.600)

So there's ways of doing that, but basically one way is to make that perturbation an expressible

Manolis Kellis (2:12:27.320)

vector so that part of your RNA reading is actually that perturbation itself.

Lex Fridman (2:12:33.160)

So you can basically put it in an expressible part so you can self drive it.

Lex Fridman (2:12:37.740)

So the point that I want to get across is that the sky's the limit.

Lex Fridman (2:12:42.120)

You basically have these tools, these building blocks of molecular biology.

Manolis Kellis (2:12:46.480)

We have these massive data sets of computational biology.

Manolis Kellis (2:12:50.280)

We have this huge ability to sort of use machine learning and statistical methods and, you

Manolis Kellis (2:12:56.160)

know, linear algebra to sort of reduce the dimensionality of all these massive data sets.

Lex Fridman (2:13:01.880)

And then you end up with a series of actionable targets that you can then couple with pharma

Lex Fridman (2:13:10.960)

and just go after systematically.

Lex Fridman (2:13:13.380)

So the ability to sort of bring genetics to the epigenomics, to the transcriptomics, to

Manolis Kellis (2:13:19.760)

the cellular readouts using these sort of high throughput perturbation technologies

Manolis Kellis (2:13:24.280)

that I'm talking about and ultimately to the organismal through the electronic health record

Manolis Kellis (2:13:30.040)

endophenotypes and ultimately the disease battery of assays at the cognitive level,

Lex Fridman (2:13:36.520)

at the physiological level and, you know, every other level.

Manolis Kellis (2:13:42.000)

There is no better or more exciting field, in my view, to be a computer scientist then

Lex Fridman (2:13:46.760)

or to be a scientist in period.

Manolis Kellis (2:13:48.640)

Basically this confluence of technologies, of computation, of data, of insight and of

Lex Fridman (2:13:54.280)

tools for manipulation is unprecedented in human history.

Lex Fridman (2:13:58.860)

And I think this is what's shaping the next century to really be a transformative century

Lex Fridman (2:14:04.620)

for our species and for our planet.

Lex Fridman (2:14:09.440)

Do you think the 21st century will be remembered for the big leaps in understanding and alleviation

Lex Fridman (2:14:17.200)

of biology?

Manolis Kellis (2:14:18.800)

If you look at the path between discovery and therapeutics, it's been on the order of

Manolis Kellis (2:14:23.720)

50 years, it's been shortened to 40, 30, 20, and now it's on the order of 10 years.

Lex Fridman (2:14:29.660)

But the huge number of technologies that are going on right now for discovery will result

Manolis Kellis (2:14:36.400)

undoubtedly in the most dramatic manipulation of human biology that we've ever seen in the

Manolis Kellis (2:14:42.600)

history of humanity in the next few years.

Lex Fridman (2:14:45.240)

Do you think we might be able to cure some of the diseases we started this conversation

Lex Fridman (2:14:48.920)

with?

Lex Fridman (2:14:49.920)

Absolutely.

Manolis Kellis (2:14:50.920)

Absolutely.

Lex Fridman (2:14:51.920)

It's only a matter of time.

Manolis Kellis (2:14:54.320)

Basically the complexity is enormous and I don't want to underestimate the complexity

Lex Fridman (2:14:58.480)

but the number of insights is unprecedented and the ability to manipulate is unprecedented

Lex Fridman (2:15:03.800)

and the ability to deliver these small molecules and other non traditional medicine perturbations,

Manolis Kellis (2:15:11.040)

there's a new generation of perturbations that you can use at the DNA level, at the

Manolis Kellis (2:15:17.440)

RNA level, at the micro RNA level, at the epigenomic level, there's a battery of new

Lex Fridman (2:15:24.440)

generations of perturbations.

Manolis Kellis (2:15:26.560)

If you couple that with cell type identifiers that can basically sense when you are in the

Manolis Kellis (2:15:32.120)

right cell based on the specific combination and then turn on that intervention for that

Manolis Kellis (2:15:36.840)

cell, you can now think of combinatorial interventions where you can basically sort of feed a synthetic

Manolis Kellis (2:15:42.560)

biology construct to someone that will basically do different things in different cells.

Lex Fridman (2:15:47.680)

So basically for cancer, this is one of the therapeutics that our collaborator Ron Weiss

Manolis Kellis (2:15:51.500)

is using to basically start sort of engineering the circuits that will use micro RNA sensors

Manolis Kellis (2:15:56.240)

of the environment to sort of know if you're in a tumor cell or if you're in an immune

Manolis Kellis (2:15:59.840)

cell or if you're in a stromal cell and so forth and basically turn on particular interventions

Manolis Kellis (2:16:04.180)

there.

Manolis Kellis (2:16:05.180)

You can sort of create constructs that are tuned to only the liver cells or only the

Manolis Kellis (2:16:11.080)

heart cells or only the brain cells and then have these new generations of therapeutics

Manolis Kellis (2:16:18.640)

coupled with this immense amount of knowledge on the sort of which targets to choose and

Lex Fridman (2:16:24.000)

what biological processes to measure and how to intervene.

Manolis Kellis (2:16:27.680)

My view is that disease is going to be fundamentally altered and alleviated as we go forward.

Manolis Kellis (2:16:36.400)

Next time we talk, we'll talk about the philosophical implications of that and the effect of life,

Lex Fridman (2:16:40.960)

but let's stick to biology for just a little longer.

Manolis Kellis (2:16:44.200)

We did pretty good today.

Lex Fridman (2:16:45.200)

We stuck to the science.

Lex Fridman (2:16:49.520)

What are you excited in terms of the future of this field, the technologies in your own

Manolis Kellis (2:16:56.000)

group, in your own mind, you're leading the world at MIT in the science and the engineering

Manolis Kellis (2:17:02.560)

of this work.

Lex Fridman (2:17:04.480)

So what are you excited about here?

Manolis Kellis (2:17:06.440)

I could not be more excited.

Lex Fridman (2:17:08.920)

We are one of many, many teams who are working on this.

Manolis Kellis (2:17:12.720)

In my team, the most exciting parts are, you know, many folds.

Lex Fridman (2:17:17.000)

So basically we've now assembled these battery of technologies.

Manolis Kellis (2:17:20.360)

We've assembled these massive, massive data sets and now we're really sort of in the stage

Lex Fridman (2:17:24.960)

of our team's path of generating disease insights.

Lex Fridman (2:17:30.460)

So we are simultaneously working on a paper on schizophrenia right now that is basically

Manolis Kellis (2:17:36.480)

using the single cell profiling technologies, using this editing and manipulation technologies

Manolis Kellis (2:17:40.880)

to basically show how the master regulators underlying changes in the brain that are sort

Manolis Kellis (2:17:47.840)

of found in schizophrenia are in fact affecting excitatory neurons and inhibitory neurons

Manolis Kellis (2:17:53.320)

in pathways that are active both in synaptic pruning, but also in early development.

Manolis Kellis (2:17:59.280)

We've basically found this set of four regulators that are connecting these two processes that

Manolis Kellis (2:18:03.220)

were previously separate in schizophrenia in sort of having a sort of more unified view

Lex Fridman (2:18:10.200)

across those two sides.

Manolis Kellis (2:18:12.720)

The second one is in the area of metabolism.

Manolis Kellis (2:18:15.520)

We basically now have a beautiful collaboration with the Goodyear lab that's basically looking

Manolis Kellis (2:18:19.280)

at multi tissue perturbations in six or seven different tissues across the body in the context

Manolis Kellis (2:18:29.160)

of exercise and in the context of nutritional interventions using both mouse and human,

Manolis Kellis (2:18:35.920)

where we can basically see what are the cell to cell communications that are changing across

Lex Fridman (2:18:41.680)

them.

Lex Fridman (2:18:42.680)

And what we're finding is this immense role of both immune cells as well as adipocyte

Manolis Kellis (2:18:47.840)

stem cells in sort of reshaping that circuitry of all of these different tissues and that's

Manolis Kellis (2:18:53.080)

sort of painting to a new path for therapeutical intervention there.

Manolis Kellis (2:18:56.920)

In Alzheimer's, it's this huge focus on microglia and now we're discovering different classes

Manolis Kellis (2:19:02.540)

of microglial cells that are basically either synaptic or immune.

Lex Fridman (2:19:10.360)

And these are playing vastly different roles in Alzheimer's versus in schizophrenia.

Lex Fridman (2:19:16.120)

And what we're finding is this immense complexity as you go further and further down of how

Manolis Kellis (2:19:22.400)

in fact there's 10 different types of microglia, each with their own sort of expression programs.

Manolis Kellis (2:19:28.400)

We used to think of them as, oh yeah, they're microglia, but in fact now we're realizing

Manolis Kellis (2:19:32.480)

just even in that sort of least abundant of cell types, there's this incredible diversity

Manolis Kellis (2:19:37.960)

there.

Lex Fridman (2:19:39.620)

The differences between brain regions is another sort of major, major insight.

Manolis Kellis (2:19:44.280)

Often one would think that, oh, astrocytes are astrocytes no matter where they are.

Lex Fridman (2:19:48.800)

But no, there's incredible region specific differences in the expression patterns of

Manolis Kellis (2:19:54.240)

all of the major brain cell types across different brain regions.

Lex Fridman (2:19:57.480)

So basically there's the neocortical regions that are sort of the recent innovation that

Manolis Kellis (2:20:01.080)

makes us so different from all other species.

Manolis Kellis (2:20:03.620)

There's the sort of reptilian brain sort of regions that are sort of much more very extremely

Manolis Kellis (2:20:10.080)

distinct.

Lex Fridman (2:20:11.080)

There's the cerebellum.

Manolis Kellis (2:20:12.080)

Each of those basically is associated in a different way with disease.

Lex Fridman (2:20:17.520)

And what we're doing now is looking into pseudo temporal models for how disease progresses

Manolis Kellis (2:20:23.680)

across different regions of the brain.

Manolis Kellis (2:20:25.820)

If you look at Alzheimer's, it basically starts in this small region called the entorhinal

Manolis Kellis (2:20:30.000)

cortex and then it spreads through the brain and through the hippocampus and ultimately

Lex Fridman (2:20:38.440)

affecting the neocortex.

Lex Fridman (2:20:39.520)

And with every brain region that it hits, it basically has a different impact on the

Manolis Kellis (2:20:46.080)

cognitive and memory aspects, orientation, short term memory, long term memory, et cetera,

Manolis Kellis (2:20:52.920)

which is dramatically affecting the cognitive path that the individuals go through.

Lex Fridman (2:20:58.320)

So what we're doing now is creating these computational models for ordering the cells

Lex Fridman (2:21:04.600)

and the regions and the individuals according to their ability to predict Alzheimer's disease.

Lex Fridman (2:21:10.560)

So we can have a cell level predictor of pathology that allows us to now create a temporal time

Manolis Kellis (2:21:17.820)

course that tells us when every gene turns on along this pathology progression and then

Manolis Kellis (2:21:22.860)

trace that across regions and pathological measures that are region specific, but also

Manolis Kellis (2:21:28.040)

cognitive measures and so on and so forth.

Lex Fridman (2:21:30.380)

So that allows us to now sort of for the first time, look at can we actually do early intervention

Manolis Kellis (2:21:35.540)

for Alzheimer's where we know that the disease starts manifesting for 10 years before you

Lex Fridman (2:21:40.920)

actually have your first cognitive loss.

Manolis Kellis (2:21:44.280)

Can we start seeing that path to build new diagnostics, new prognostics, new biomarkers

Lex Fridman (2:21:50.360)

for this sort of early intervention in Alzheimer's?

Manolis Kellis (2:21:54.420)

The other aspect that we're looking at is mosaicism.

Manolis Kellis (2:21:57.080)

We talked about the common variants and the rare variants, but in addition to those rare

Manolis Kellis (2:22:01.120)

variants as your initial cell that forms the zygote divides and divides and divides, with

Lex Fridman (2:22:08.520)

every cell division there are additional mutations that are happening.

Lex Fridman (2:22:12.480)

So what you end up with is your brain being a mosaic of multiple different types of genetic

Lex Fridman (2:22:18.000)

underpinnings.

Manolis Kellis (2:22:19.320)

Some cells contain a mutation that other cells don't have.

Lex Fridman (2:22:23.380)

So every human has the common variants that all of us carry to some degree, the rare variants

Manolis Kellis (2:22:31.200)

that your immediate tree of the human species carries, and then there's the somatic variant,

Manolis Kellis (2:22:37.360)

which is the tree that happened after the zygote that sort of forms your own body.

Lex Fridman (2:22:44.280)

So these somatic alterations is something that has been previously inaccessible to study

Lex Fridman (2:22:50.840)

in human postmortem samples.

Lex Fridman (2:22:53.240)

But right now with the advent of single cell RNA sequencing, in this particular case, we're

Manolis Kellis (2:22:58.240)

using the well based sequencing, which is much more expensive, but gives you a lot richer

Manolis Kellis (2:23:01.920)

information about each of those transcripts.

Lex Fridman (2:23:04.560)

So we're using now that richer information to infer mutations that have happened in each

Manolis Kellis (2:23:10.200)

of the thousands of genes that sort of are active in these cells, and then understand

Lex Fridman (2:23:16.640)

how the genome relates to the function, this genotype phenotype relationship that we usually

Manolis Kellis (2:23:25.320)

build in GWAS between in genome wide association studies between genetic variation and disease.

Manolis Kellis (2:23:31.400)

We're now building that at the cell level, where for every cell, we can relate the unique

Manolis Kellis (2:23:36.400)

specific genome of that cell with the expression patterns of that cell, and the predicted function

Manolis Kellis (2:23:42.920)

using these predictive models that I mentioned before on this regulation for cognition for

Manolis Kellis (2:23:47.480)

pathology in Alzheimer's at the cell level.

Lex Fridman (2:23:51.000)

And what we're finding is that the genes that are altered and the genetic regions that are

Manolis Kellis (2:23:54.960)

altered in common variants versus rare variants versus somatic variants are actually very

Lex Fridman (2:23:59.640)

different from each other.

Manolis Kellis (2:24:01.280)

The somatic variants are pointing to neuronal energetics and oligodendrocyte functions that

Manolis Kellis (2:24:08.860)

are not visible in the genetic legions that you find for the common variants, probably

Manolis Kellis (2:24:13.000)

because they have too strong of an effect that evolution is just not tolerating them

Lex Fridman (2:24:17.480)

on the common side of the allele frequency spectrum.

Lex Fridman (2:24:20.960)

So the somatic one, that's the variation that happens after the zygote, after you individual.

Manolis Kellis (2:24:26.360)

I mean, this is a dumb question, but there's mutation and variation, I guess that happens

Manolis Kellis (2:24:31.600)

there.

Lex Fridman (2:24:32.600)

And you're saying that they're through this, if we focus in on individual cells, we're

Manolis Kellis (2:24:37.200)

able to detect the story that's interesting there, and that might be a very unique kind

Manolis Kellis (2:24:42.640)

of important variability that arises for, you said neuronal or something that would

Manolis Kellis (2:24:49.320)

sound...

Lex Fridman (2:24:50.320)

Energetics.

Manolis Kellis (2:24:51.320)

Energetics, sounds like a cool term.

Manolis Kellis (2:24:52.320)

So, I mean, the metabolism of humans is dramatically altered from that of nearby species.

Manolis Kellis (2:24:59.520)

We talked about that last time that basically we are able to consume meat that is incredibly

Manolis Kellis (2:25:04.500)

energy rich, and that allows us to sort of have functions that are meeting this humongous

Manolis Kellis (2:25:13.240)

brain that we have.

Lex Fridman (2:25:14.240)

So basically on one hand, every one of our brain cells is much more energy efficient

Manolis Kellis (2:25:18.280)

than our neighbors, than our relatives.

Lex Fridman (2:25:20.560)

Number two, we have way more of these cells.

Lex Fridman (2:25:23.360)

And number three, we have this new diet that allows us to now feed all these needs.

Manolis Kellis (2:25:30.260)

That basically creates a massive amount of damage, oxidative damage from this huge super

Manolis Kellis (2:25:36.540)

powered factory of ideas and thoughts that we carry in our skull.

Lex Fridman (2:25:42.360)

And that factory has energetic needs, and there's a lot of sort of biological processes

Manolis Kellis (2:25:47.540)

underlying that, that we are finding are altered in the context of Alzheimer's disease.

Lex Fridman (2:25:52.960)

That's fascinating.

Lex Fridman (2:25:53.960)

So you have to consider all of these systems if you want to understand even something like

Manolis Kellis (2:25:59.680)

diseases that you would maybe traditionally associate with just the particular cells of

Manolis Kellis (2:26:04.440)

the brain.

Lex Fridman (2:26:07.440)

The immune system, the metabolic system, the metabolic system.

Lex Fridman (2:26:11.240)

And these are all the things that makes us uniquely human.

Lex Fridman (2:26:13.440)

So our immune system is dramatically different from that of our neighbors.

Manolis Kellis (2:26:17.120)

Our societies are so much more clustered.

Manolis Kellis (2:26:19.600)

The history of infection that have plagued the human population is dramatically different

Manolis Kellis (2:26:24.840)

from every other species.

Manolis Kellis (2:26:27.080)

The way that our society and our population has sort of exploded has basically put unique

Manolis Kellis (2:26:31.320)

pressures on our immune system.

Lex Fridman (2:26:33.360)

And our immune system has both coped with that density and also been shaped by, as I

Manolis Kellis (2:26:37.480)

mentioned, the vast amount of death that has happened in the Black Plague and other sort

Lex Fridman (2:26:42.200)

of selective events in human history, famines, ice ages, and so forth.

Lex Fridman (2:26:47.180)

So that's number one on the sort of immune side.

Lex Fridman (2:26:49.940)

On the metabolic side, again, we are able to sort of run marathons.

Manolis Kellis (2:26:55.560)

I don't know if you remember the sort of human versus horse experiment where the horse actually

Lex Fridman (2:26:59.040)

tires out faster than the human and the human actually wins.

Lex Fridman (2:27:03.480)

So on the metabolic side, we're dramatically different.

Lex Fridman (2:27:05.940)

On the immune side, we're dramatically different.

Manolis Kellis (2:27:07.560)

On the brain side, again, you know, no need to sort of, you know, it's a no brainer of

Lex Fridman (2:27:12.400)

how our brain is like just enormously more capable.

Lex Fridman (2:27:16.880)

And then, you know, in the side of cancer, so basically the cancers that humans are having,

Lex Fridman (2:27:21.740)

the exposures, the environmental exposures is again, dramatically different.

Lex Fridman (2:27:25.940)

And the lifespan, the expansion of human lifespan is unseen in any other species in, you know,

Lex Fridman (2:27:32.880)

recent evolutionary history.

Lex Fridman (2:27:35.720)

And that now leads to a lot of new disorders that are starting to, you know, manifest late

Lex Fridman (2:27:42.360)

in life.

Lex Fridman (2:27:43.920)

So you know, Alzheimer's is one example where basically, you know, these vast energetic

Manolis Kellis (2:27:48.200)

needs over a lifetime of thinking can basically lead to all of these debris and eventually

Manolis Kellis (2:27:54.800)

saturate the system and lead to, you know, Alzheimer's in the late life.

Lex Fridman (2:28:00.840)

But there's, you know, there's just such a dramatic set of frontiers when it comes to

Manolis Kellis (2:28:07.440)

aging research that, you know, so what I often like to say is that if you want to engineer

Lex Fridman (2:28:14.360)

a car to go from 70 miles an hour to 120 miles an hour, that's fine.

Manolis Kellis (2:28:18.240)

You can basically, you know, fix a few components.

Manolis Kellis (2:28:20.480)

If you wanted to now go at 400 miles an hour, you have to completely redesign the entire

Manolis Kellis (2:28:24.240)

car because the system has just not evolved to go that far.

Manolis Kellis (2:28:31.220)

Basically our human body has only evolved to live to, I don't know, 120, maybe we can

Manolis Kellis (2:28:36.480)

get to 150 with minor changes.

Lex Fridman (2:28:39.280)

But if, you know, as we start pushing these frontiers for not just living, but well living,

Manolis Kellis (2:28:45.240)

the Fzine that we talked about last time.

Lex Fridman (2:28:48.240)

So to basically push Fzine into the 80s and 90s and a hundreds and, you know, much further

Manolis Kellis (2:28:53.200)

than that, we will face new challenges that have, you know, never been faced before in

Manolis Kellis (2:29:00.400)

terms of cancer, the number of divisions, in terms of Alzheimer's and brain related

Manolis Kellis (2:29:04.200)

disorders, in terms of metabolic disorders, in terms of regeneration, there's just so

Lex Fridman (2:29:08.880)

many different frontiers ahead of us.

Lex Fridman (2:29:10.920)

So I am thrilled about where we're heading.

Lex Fridman (2:29:14.040)

So basically I see this confluence in my lab and many other labs of AI, of, you know, sort

Manolis Kellis (2:29:20.600)

of, you know, the next frontier of AI for drug design.

Lex Fridman (2:29:22.920)

So basically these sort of graph neural networks on specific chemical designs that allow you

Manolis Kellis (2:29:30.520)

to create new generations of therapeutics.

Manolis Kellis (2:29:34.840)

These molecular biology tricks for intervening at the system at every level, these personalized

Manolis Kellis (2:29:42.400)

medicine prediction, diagnosis, and prognosis using the electronic health records and using

Manolis Kellis (2:29:49.960)

these polygenic risk scores weighted by the burden, the number of mutations that are accumulating

Manolis Kellis (2:29:56.640)

across common rare and somatic variants, the burden converging across all of these different

Manolis Kellis (2:30:03.340)

molecular pathways, the delivery of specific drugs and specific interventions into specific

Manolis Kellis (2:30:10.000)

cell types.

Lex Fridman (2:30:11.000)

And again, you've talked with Bob Langer about this, there's, you know, many giants in that

Manolis Kellis (2:30:14.080)

field.

Lex Fridman (2:30:15.080)

And then the last concept is not intervening at the single gene level.

Manolis Kellis (2:30:20.560)

I want you to sort of conceptualize the concept of an on target side effect.

Lex Fridman (2:30:27.600)

What is an on target side effect?

Manolis Kellis (2:30:29.200)

An off target side effect is when you design a molecule to target one gene and instead

Lex Fridman (2:30:33.320)

it targets another gene and you have side effects because of that.

Lex Fridman (2:30:36.720)

And on target side effect is when your molecule does exactly what you were expecting, but

Lex Fridman (2:30:41.040)

that gene is plyotropic.

Manolis Kellis (2:30:43.840)

Plyo means many, tropos means ways, many ways, it acts in many ways.

Lex Fridman (2:30:48.160)

It's a multifunctional gene.

Lex Fridman (2:30:50.040)

So you find that this gene plays a role in this, but as we talked about the wiring of

Lex Fridman (2:30:55.320)

genes to phenotypes is extremely dense and extremely complex.

Lex Fridman (2:30:59.000)

So the next stage of intervention will be intervening not at the gene level, but at

Lex Fridman (2:31:04.000)

the network level.

Manolis Kellis (2:31:06.160)

Intervening at the set of pathways and the set of genes with multi input perturbations

Manolis Kellis (2:31:11.440)

to the system, multi input modulations, pharmaceutical or other interventional, and that basically

Manolis Kellis (2:31:18.040)

allow you to now work at the sort of full level of understanding, not just in your brain,

Lex Fridman (2:31:24.980)

but across your body, not just in one gene, but across the set of pathways and so on and

Lex Fridman (2:31:29.400)

so forth for every one of these disorders.

Lex Fridman (2:31:31.980)

So I think that we're finally at the level of systems medicine of basically instead of

Manolis Kellis (2:31:37.320)

sort of medicine being at the single gene level, medicine being at the systems level

Manolis Kellis (2:31:42.120)

where it can be personalized based on the specific set of genetic markers and genetic

Manolis Kellis (2:31:46.480)

perturbations that you are either born with or that you have developed during your lifetime.

Manolis Kellis (2:31:53.040)

Your unique set of exposures, your unique set of biomarkers, and your unique set of

Manolis Kellis (2:31:59.480)

current set of conditions through your EHR and other ways.

Lex Fridman (2:32:06.480)

And the precision component of intervening extremely precisely in the specific pathways

Lex Fridman (2:32:12.920)

and the specific combinations of genes that should be modulated to sort of bring you from

Manolis Kellis (2:32:16.840)

the disease state to the physiologically normal state or even to physiologically improved

Manolis Kellis (2:32:23.480)

state through this combination of interventions.

Lex Fridman (2:32:25.640)

So that's in my view, the field where basically computer science comes together with artificial

Manolis Kellis (2:32:30.080)

intelligence statistics, all of these other tools, molecular biology technologies and

Manolis Kellis (2:32:34.200)

biotechnology and pharmaceutical technologies that are sort of revolutionary in the way

Manolis Kellis (2:32:37.960)

of intervention.

Lex Fridman (2:32:38.960)

And of course, this massive amount of molecular biology and data gathering and generation

Lex Fridman (2:32:43.240)

and perturbation in massively parallel ways.

Lex Fridman (2:32:46.360)

So there's no better way.

Manolis Kellis (2:32:47.700)

There's no better time.

Manolis Kellis (2:32:49.740)

There's no better place to be sort of looking at this whole confluence of ideas.

Lex Fridman (2:32:56.800)

And I'm just so thrilled to be a small part of this amazing, enormous ecosystem.

Manolis Kellis (2:33:01.440)

It's exciting to imagine what humans of 100, 200 years from now, what their life experience

Manolis Kellis (2:33:07.520)

is like, because these ideas seem to have potential to transform the quality of life

Manolis Kellis (2:33:13.720)

that, when they look back at us, they probably wonder how we were put up with all the suffering

Manolis Kellis (2:33:22.200)

in the world.

Lex Fridman (2:33:23.200)

Manolis, it's a huge honor.

Manolis Kellis (2:33:25.480)

Thank you for spending this early Sunday morning with me.

Lex Fridman (2:33:29.240)

I deeply appreciate it.

Manolis Kellis (2:33:30.240)

See you next time.

Lex Fridman (2:33:31.240)

Sounds like a plan.

Manolis Kellis (2:33:32.240)

Thank you, Lex.

Lex Fridman (2:33:33.960)

Thanks for listening to this conversation with Manolis Kellis.

Lex Fridman (2:33:36.880)

And thank you to our sponsors, SEMrush, which is an SEO optimization tool.

Lex Fridman (2:33:43.280)

Pessimist Archive, which is one of my favorite history podcasts.

Manolis Kellis (2:33:47.400)

8Sleep, which is a self cooling mattress with smart sensors and an app.

Lex Fridman (2:33:52.680)

And finally, BetterHelp, which is an online therapy service.

Manolis Kellis (2:33:57.360)

Please check out these sponsors in the description to get a discount and to support this podcast.

Manolis Kellis (2:34:02.520)

If you enjoy this thing, subscribe on YouTube, review it with 5 Stars and Apple Podcasts,

Manolis Kellis (2:34:08.000)

follow on Spotify, support on Patreon, or connect with me on Twitter at Lex Friedman.

Lex Fridman (2:34:13.240)

And now, let me leave you with some words from Haruki Murakami.

Manolis Kellis (2:34:19.120)

Human beings are ultimately nothing but carriers, passageways for genes.

Lex Fridman (2:34:24.580)

They ride us into the ground like racehorses from generation to generation.

Manolis Kellis (2:34:30.300)

Genes don't think about what constitutes good or evil.

Lex Fridman (2:34:34.160)

They don't care whether we're happy or unhappy.

Manolis Kellis (2:34:37.040)

We're just means to an end for them.

Lex Fridman (2:34:40.060)

The only thing they think about is what is most efficient for them.

Manolis Kellis (2:34:45.960)

Thank you for listening, and hope to see you next time.

Lex Fridman (30:05.480)

to solve, which aren't.

Manolis Kellis (30:07.280)

I love your question because he puts it in the context of a global effort rather than

Lex Fridman (30:13.720)

just the local effort.

Lex Fridman (30:15.120)

So basically if you look at the global aspect, exercise and nutrition are two interventions

Lex Fridman (30:22.560)

that we can as a society make a much better job at.

Lex Fridman (30:27.040)

So if you think about sort of the availability of cheap food, it's extremely high in calories.

Lex Fridman (30:33.080)

It's extremely detrimental for you, like a lot of processed food, et cetera.

Lex Fridman (30:36.960)

So if we change that equation and as a society, we made availability of healthy food much,

Manolis Kellis (30:43.120)

much easier and charged a burger at McDonald's, the price that it costs on the health system,

Manolis Kellis (30:52.520)

then people would actually start buying more healthy foods.

Lex Fridman (30:56.360)

So basically that's sort of a societal intervention, if you wish.

Manolis Kellis (30:59.600)

In the same way, increasing empathy, increasing education, increasing the social framework

Lex Fridman (31:06.580)

and support would basically lead to fewer suicides.

Manolis Kellis (31:10.040)

It would lead to fewer murders.

Lex Fridman (31:11.880)

It would lead to fewer deaths overall.

Lex Fridman (31:15.880)

So that's something that we as a society can do.

Lex Fridman (31:19.100)

You can also think about external factors versus internal factors.

Lex Fridman (31:21.940)

So the external factors are basically communicable diseases like COVID, like the flu, et cetera.

Lex Fridman (31:27.440)

And the internal factors are basically things like cancer and Alzheimer's where basically

Manolis Kellis (31:33.200)

your genetics will eventually drive you there.

Lex Fridman (31:38.560)

And then of course, with all of these factors, every single disease has both the genetic

Manolis Kellis (31:43.520)

component and environmental component.

Lex Fridman (31:46.200)

So heart disease, huge genetic contribution, Alzheimer's, it's like 60% plus genetic.

Lex Fridman (31:55.700)

So I think it's like 79% heritability.

Lex Fridman (31:59.040)

So that basically means that genetics alone explains 79% of Alzheimer's incidents.

Lex Fridman (32:06.400)

And yes, there's a 21% environmental component where you could basically enrich your cognitive

Manolis Kellis (32:14.040)

environment, enrich your social interactions, read more books, learn a foreign language,

Manolis Kellis (32:21.040)

go running, you know, sort of have a more fulfilling life.

Manolis Kellis (32:24.800)

All of that will actually decrease Alzheimer's, but there's a limit to how much that can impact

Manolis Kellis (32:29.320)

because of the huge genetic footprint.

Lex Fridman (32:31.240)

So this is fascinating.

Lex Fridman (32:32.240)

So each one of these problems have a genetic component and an environment component.

Lex Fridman (32:38.860)

And so like when there's a genetic component, what can we do about some of these diseases?

Lex Fridman (32:43.520)

And have you worked on what can you say that's in terms of problems that are solvable here

Lex Fridman (32:48.520)

or understandable?

Lex Fridman (32:50.520)

So my group works on the genetic component, but I would argue that understanding the genetic

Lex Fridman (32:55.740)

component can have a huge impact even on the environmental component.

Lex Fridman (32:59.700)

Why is that?

Lex Fridman (33:00.800)

Because genetics gives us access to mechanism.

Lex Fridman (33:03.560)

And if we can alter the mechanism, if we can impact the mechanism, we can perhaps counteract

Lex Fridman (33:09.580)

some of the environmental components.

Lex Fridman (33:12.080)

So understanding the biological mechanisms leading to disease is extremely important

Lex Fridman (33:18.240)

in being able to intervene.

Lex Fridman (33:20.820)

But when you can intervene and what, you know, the analogy that I like to give is for example,

Lex Fridman (33:26.040)

for obesity, you know, think of it as a giant bathtub of fat.

Manolis Kellis (33:29.880)

There's basically fat coming in from your diet and there's fat coming out from your

Lex Fridman (33:35.800)

exercise.

Manolis Kellis (33:36.800)

Okay.

Lex Fridman (33:37.800)

So that's an in out equation and that's the equation that everybody's focusing on.

Lex Fridman (33:42.240)

But your metabolism impacts that, you know, bathtub.

Lex Fridman (33:47.740)

Basically your metabolism controls the rate at which you're burning energy.

Manolis Kellis (33:53.080)

It controls the rate at which you're storing energy.

Lex Fridman (33:56.640)

And it also teaches you about the various valves that control the input and the output

Manolis Kellis (34:02.800)

equation.

Lex Fridman (34:04.020)

So if we can learn from the genetics, the valves, we can then manipulate those valves.

Lex Fridman (34:11.320)

And even if the environment is feeding you a lot of fat and getting a little that out,

Manolis Kellis (34:16.060)

you can just poke another hole at the bathtub and just get a lot of the fat out.

Manolis Kellis (34:19.840)

Yeah, that's fascinating.

Lex Fridman (34:21.160)

Yeah.

Lex Fridman (34:22.160)

So we're not just passive observers of our genetics.

Lex Fridman (34:25.840)

The more we understand, the more we can come up with actual treatments.

Lex Fridman (34:29.640)

And I think that's an important aspect to realize when people are thinking about strong

Lex Fridman (34:35.680)

effect versus weak effect variants.

Lex Fridman (34:38.080)

So some variants have strong effects.

Manolis Kellis (34:39.580)

We talked about these Mendelian disorders where a single gene has a sufficiently large

Manolis Kellis (34:43.400)

effect, penetrance, expressivity, and so on and so forth, that basically you can trace

Manolis Kellis (34:49.420)

it in families with cases and not cases, cases, not cases, and so on and so forth.

Lex Fridman (34:55.320)

But so these are the genes that everybody says, oh, that's the genes we should go after

Lex Fridman (35:02.840)

because that's a strong effect gene.

Manolis Kellis (35:04.880)

I like to think about it slightly differently.

Manolis Kellis (35:06.860)

These are the genes where genetic impacts that have a strong effect were tolerated because

Manolis Kellis (35:15.440)

every single time we have a genetic association with disease, it depends on two things.

Lex Fridman (35:20.200)

Number one, the obvious one, whether the gene has an impact on the disease.

Manolis Kellis (35:24.680)

Number two, the more subtle one is whether there is genetic variation standing and circulating

Lex Fridman (35:32.180)

and segregating in the human population that impacts that gene.

Manolis Kellis (35:37.680)

Some genes are so darn important that if you mess with them, even a tiny little amount,

Lex Fridman (35:44.480)

that person's dead.

Lex Fridman (35:46.440)

So those genes don't have variation.

Lex Fridman (35:49.020)

You're not going to find a genetic association if you don't have variation.

Manolis Kellis (35:53.040)

That doesn't mean that the gene has no role.

Lex Fridman (35:55.400)

It simply means that the gene tolerates no mutations.

Lex Fridman (35:59.120)

So that's actually a strong signal when there's no variation.

Lex Fridman (36:01.480)

That's so fascinating.

Manolis Kellis (36:02.480)

Exactly.

Lex Fridman (36:03.480)

Genes that have very little variation are hugely important.

Manolis Kellis (36:06.780)

You can actually rank the importance of genes based on how little variation they have.

Lex Fridman (36:10.840)

And those genes that have very little variation but no association with disease, that's a

Manolis Kellis (36:16.920)

very good metric to say, oh, that's probably a developmental gene because we're not good

Lex Fridman (36:20.440)

at measuring those phenotypes.

Lex Fridman (36:22.840)

So it's genes that you can tell evolution has excluded mutations from, but yet we can't

Lex Fridman (36:29.040)

see them associated with anything that we can measure nowadays.

Manolis Kellis (36:32.120)

It's probably early embryonic lethal.

Lex Fridman (36:34.840)

What are all the words you just said?

Lex Fridman (36:36.200)

Early embryonic what?

Lex Fridman (36:37.760)

Lethal.

Lex Fridman (36:38.760)

Meaning?

Lex Fridman (36:39.760)

Meaning that that embryo will die.

Manolis Kellis (36:40.760)

Okay.

Manolis Kellis (36:41.760)

There's a bunch of stuff that is required for a stable functional organism across the

Manolis Kellis (36:49.160)

board for an entire species, I guess.

Lex Fridman (36:53.880)

If you look at sperm, it expresses thousands of proteins.

Lex Fridman (36:58.680)

Does sperm actually need thousands of proteins?

Lex Fridman (37:01.240)

No, but it's probably just testing them.

Lex Fridman (37:05.320)

So my speculation is that misfolding of these proteins is an early test for failure.

Lex Fridman (37:11.960)

So that out of the millions of sperm that are possible, you select the subset that are

Manolis Kellis (37:18.440)

just not grossly misfolding thousands of proteins.

Lex Fridman (37:21.920)

So it's kind of an assert that this is folded correctly.

Manolis Kellis (37:25.720)

Correct.

Lex Fridman (37:26.720)

Yeah.

Manolis Kellis (37:27.720)

This just because if this little thing about the folding of a protein isn't correct, that

Lex Fridman (37:32.560)

probably means somewhere down the line, there's a bigger issue.

Manolis Kellis (37:35.720)

That's exactly right.

Lex Fridman (37:36.720)

So fail fast.

Lex Fridman (37:37.720)

So basically if you look at the mammalian investment in a newborn, that investment is

Lex Fridman (37:45.100)

enormous in terms of resources.

Lex Fridman (37:47.720)

So mammals have basically evolved mechanisms for fail fast.

Manolis Kellis (37:52.840)

Where basically in those early months of development, I mean it's horrendous of course at the personal

Manolis Kellis (37:58.880)

level when you lose your future child, but in some ways there's so little hope for that

Manolis Kellis (38:08.680)

child to develop and sort of make it through the remaining months that sort of fail fast

Manolis Kellis (38:12.880)

is probably a good evolutionary principle for mammals.

Lex Fridman (38:19.560)

And of course humans have a lot of medical resources that you can sort of give those

Manolis Kellis (38:24.920)

children a chance and we have so much more success in sort of giving folks who have these

Manolis Kellis (38:33.120)

strong carrier mutations a chance, but if they're not even making it through the first

Manolis Kellis (38:37.040)

three months, we're not going to see them.

Lex Fridman (38:39.860)

So that's why when we say what are the most important genes to focus on, the ones that

Manolis Kellis (38:45.080)

have a strong effect mutation or the ones that have a weak effect mutation, well the

Manolis Kellis (38:50.040)

jury might be out because the ones that have a strong effect mutation are basically not

Manolis Kellis (38:57.080)

mattering as much.

Manolis Kellis (38:58.720)

The ones that only have weak effect mutations by understanding through genetics that they

Manolis Kellis (39:04.960)

have a weak effect mutation and understanding that they have a causal role on the disease,

Manolis Kellis (39:10.200)

we can then say, okay, great, evolution has only tolerated a 2% change in that gene.

Manolis Kellis (39:15.720)

Pharmaceutically I can go in and induce a 70% change in that gene and maybe I will poke

Manolis Kellis (39:22.560)

another hole at the bathtub that was not easy to control in many of the other sort of strong

Manolis Kellis (39:33.220)

effect genetic variants.

Lex Fridman (39:35.160)

So there's this beautiful map of across the population of things that you're saying strong

Lex Fridman (39:41.800)

and weak effects, so stuff with a lot of mutations and stuff with little mutations with no mutations

Lex Fridman (39:48.200)

and you have this map and it lays out the puzzle.

Manolis Kellis (39:51.360)

Yeah.

Lex Fridman (39:52.360)

So when I say strong effect, I mean at the level of individual mutations.

Lex Fridman (39:56.120)

So basically genes where, so you have to think of first the effect of the gene on the disease.

Manolis Kellis (40:03.640)

Remember how I was sort of painting that map earlier from genetics all the way to phenotype.

Manolis Kellis (40:10.240)

That gene can have a strong effect on the disease, but the genetic variant might have

Lex Fridman (40:15.960)

a weak effect on the gene.

Lex Fridman (40:18.880)

So basically when you ask what is the effect of that genetic variant on the disease, it

Manolis Kellis (40:24.960)

could be that that genetic variant impacts the gene by a lot and then the gene impacts

Manolis Kellis (40:29.560)

the disease by a little, or it could be that the genetic variants impacts the gene by a

Lex Fridman (40:33.240)

little and then the gene impacts the disease by a lot.

Lex Fridman (40:35.880)

So what we care about is genes that impact the disease a lot, but genetics gives us the

Manolis Kellis (40:41.720)

full equation and what I would argue is if we couple the genetics with expression variation

Manolis Kellis (40:51.920)

to basically ask what genes change by a lot and which genes correlate with disease by

Manolis Kellis (41:00.400)

a lot, even if the genetic variants change them by a little, then those are the best

Manolis Kellis (41:06.200)

places to intervene.

Manolis Kellis (41:07.200)

Those are the best places where pharmaceutical, if I have even a modest effect, I will have

Manolis Kellis (41:13.200)

a strong effect on the disease, whereas those genetic variants that have a huge effect on

Manolis Kellis (41:17.120)

the disease, I might not be able to change that gene by this much without affecting all

Manolis Kellis (41:21.120)

kinds of other things.

Lex Fridman (41:22.360)

Interesting.

Lex Fridman (41:23.360)

So that's what we're looking at.

Lex Fridman (41:26.040)

What have we been able to find in terms of which disease could be helped?

Manolis Kellis (41:31.800)

Again, don't get me started.

Lex Fridman (41:37.280)

We have found so much.

Manolis Kellis (41:38.960)

Our understanding of disease has changed so dramatically with genetics.

Lex Fridman (41:46.000)

I mean places that we had no idea would be involved.

Lex Fridman (41:49.000)

So one of the worst things about my genome is that I have a genetic predisposition to

Lex Fridman (41:53.920)

age related macular degeneration, AMD.

Lex Fridman (41:56.520)

So it's a form of blindness that causes you to lose the central part of your vision progressively

Lex Fridman (42:02.260)

as you grow older.

Manolis Kellis (42:04.400)

My increased risk is fairly small.

Lex Fridman (42:06.240)

I have an 8% chance.

Manolis Kellis (42:07.680)

You only have a 6% chance.

Lex Fridman (42:10.080)

I'm an average.

Manolis Kellis (42:11.080)

By the way, when you say my, you mean literally yours.

Lex Fridman (42:14.560)

You know this about you.

Manolis Kellis (42:15.880)

I know this about me.

Manolis Kellis (42:18.000)

Which is kind of, I mean philosophically speaking is a pretty powerful thing to live with.

Manolis Kellis (42:26.500)

Maybe that's, so we agreed to talk again by the way for the listeners to where we're going

Lex Fridman (42:31.680)

to try to focus on science today and a little bit of philosophy next time.

Lex Fridman (42:36.080)

But it's interesting to think about the more you're able to know about yourself from the

Manolis Kellis (42:42.880)

genetic information in terms of the diseases, how that changes your own view of life.

Lex Fridman (42:49.360)

So there's a lot of impact there and there's something called genetics exceptionalism,

Manolis Kellis (42:56.000)

which basically thinks of genetics as something very, very different than everything else

Manolis Kellis (43:01.040)

as a type of determinism.

Lex Fridman (43:04.200)

And you know, let's talk about that next time.

Lex Fridman (43:07.320)

So basically.

Lex Fridman (43:08.320)

That's a good preview.

Manolis Kellis (43:09.320)

Yeah.

Lex Fridman (43:10.320)

So let's go back to AMD.

Lex Fridman (43:11.680)

So basically with AMD, we have no idea what causes AMD.

Lex Fridman (43:16.920)

You know, it was, it was a mystery until the genetics were worked out.

Lex Fridman (43:23.700)

And now the fact that I know that I have a predisposition allows me to sort of make some

Manolis Kellis (43:28.640)

life choices, number one, but number two, the genes that lead to that predisposition

Manolis Kellis (43:34.720)

give us insights as to how does it actually work.

Lex Fridman (43:38.520)

And that's a place where genetics gave us something totally unexpected.

Lex Fridman (43:42.960)

So there's a complement pathway, which is an immune function pathway that was in, you

Lex Fridman (43:52.000)

know, most of the loci associated with AMD.

Lex Fridman (43:55.940)

And that basically told us that, wow, there's an immune basis to this eye disorder that

Lex Fridman (44:02.600)

people had just not expected before.

Manolis Kellis (44:05.180)

If you look at complement, it was recently also implicated in schizophrenia.

Lex Fridman (44:11.160)

And there's a type of microglia that is involved in synaptic pruning.

Lex Fridman (44:17.280)

So synapses are the connections between neurons.

Lex Fridman (44:20.560)

And in this whole use it or lose it view of mental cognition and other capabilities, you

Manolis Kellis (44:27.160)

basically have microglia, which are immune cells that are sort of constantly traversing

Manolis Kellis (44:32.960)

your brain and then pruning neuronal connections, pruning synaptic connections that are not

Manolis Kellis (44:38.640)

utilized.

Lex Fridman (44:40.260)

So in schizophrenia, there's thought to be a change in the pruning that basically if

Manolis Kellis (44:47.960)

you don't prune your synapses the right way, you will actually have an increased role of

Lex Fridman (44:53.280)

schizophrenia.

Manolis Kellis (44:54.280)

This is something that was completely unexpected for schizophrenia.

Manolis Kellis (44:57.160)

Of course, we knew it has to do with neurons, but the role of the complement complex, which

Manolis Kellis (45:01.560)

is also implicated in AMD, which is now also implicated in schizophrenia, was a huge surprise.

Lex Fridman (45:06.520)

What's the complement complex?

Lex Fridman (45:08.040)

So it's basically a set of genes, the complement genes that are basically having various immune

Lex Fridman (45:13.960)

roles.

Lex Fridman (45:15.460)

And as I was saying earlier, our immune system has been coopted for many different roles

Lex Fridman (45:19.940)

across the body.

Lex Fridman (45:21.220)

So they actually play many diverse roles.

Lex Fridman (45:23.440)

And somehow the immune system is connected to the synaptic pruning process, the process.

Manolis Kellis (45:29.600)

Exactly.

Lex Fridman (45:30.600)

So the prune cells were coopted to prune synapses.

Lex Fridman (45:33.080)

How did you figure this out?

Lex Fridman (45:35.720)

How does one go about figuring this intricate connection, like pipeline of connections out?

Manolis Kellis (45:41.920)

Yeah.

Lex Fridman (45:42.920)

Let me give you another example.

Lex Fridman (45:44.280)

So Alzheimer's disease, the first place that you would expect it to act is obviously the

Lex Fridman (45:48.920)

brain.

Lex Fridman (45:49.920)

So we had basically this roadmap epigenomics consortium view of the human epigenome, the

Manolis Kellis (45:57.000)

largest map of the human epigenome that has ever been built across 127 different tissues

Lex Fridman (46:04.440)

and samples with dozens of epigenomic marks measured in hundreds of donors.

Lex Fridman (46:10.560)

So what we've basically learned through that is that you basically can map what are the

Manolis Kellis (46:16.600)

active gene regulatory elements for every one of the tissues in the body.

Lex Fridman (46:20.280)

And then we connected these gene regulatory active maps of basically what regions of the

Manolis Kellis (46:27.400)

human genome are turning on in every one of different tissues.

Manolis Kellis (46:32.000)

We then can go back and say, where are all of the genetic loci that are associated with

Lex Fridman (46:38.600)

disease?

Manolis Kellis (46:39.600)

This is something that my group, I think was the first to do back in 2010 in this Ernst

Manolis Kellis (46:46.400)

Nature Biotech paper, but basically we were for the first time able to show that specific

Manolis Kellis (46:52.040)

chromatin states, specific epigenomic states, in that case enhancers, were in fact enriched

Manolis Kellis (46:58.560)

in disease associated variants.

Lex Fridman (47:00.720)

We pushed that further in the Ernst Nature paper a year later.

Lex Fridman (47:05.640)

And then in this roadmap epigenomics paper a few years after that, but basically that

Manolis Kellis (47:12.680)

matrix that you mentioned earlier was in fact the first time that we could see what genetic

Manolis Kellis (47:18.160)

traits have genetic variants that are enriched in what tissues in the body.

Lex Fridman (47:26.360)

And a lot of that map made complete sense.

Manolis Kellis (47:28.800)

If you looked at a diversity of immune traits like allergies and type one diabetes and so

Manolis Kellis (47:33.800)

on and so forth, you basically could see that they were enriching, that the genetic variants

Manolis Kellis (47:38.920)

associated with those traits were enriched in enhancers in these gene regulatory elements

Manolis Kellis (47:44.680)

active in T cells and B cells and hematopoietic stem cells and so on and so forth.

Lex Fridman (47:49.280)

So that basically gave us a confirmation in many ways that those immune traits were indeed

Lex Fridman (47:56.960)

enriching immune cells.

Manolis Kellis (48:00.360)

If you looked at type two diabetes, you basically saw an enrichment in only one type of sample

Lex Fridman (48:06.080)

and it was pancreatic islets.

Lex Fridman (48:08.960)

And we know that type two diabetes sort of stems from the dysregulation of insulin in

Lex Fridman (48:14.960)

the beta cells of pancreatic islets.

Lex Fridman (48:17.440)

And that sort of was spot on, super precise.

Lex Fridman (48:21.200)

If you looked at blood pressure, where would you expect blood pressure to occur?

Manolis Kellis (48:25.880)

You know, I don't know, maybe in your metabolism and ways that you process coffee or something

Lex Fridman (48:29.880)

like that.

Manolis Kellis (48:30.880)

Maybe in your brain, the way that you stress out and increases your blood pressure, et

Lex Fridman (48:33.880)

cetera.

Lex Fridman (48:34.880)

So the blood pressure localized specifically in the left ventricle of the heart.

Lex Fridman (48:40.360)

So the enhancers of the left ventricle in the heart contained a lot of genetic variants

Manolis Kellis (48:44.100)

associated with blood pressure.

Manolis Kellis (48:46.760)

If you look at height, we found an enrichment specifically in embryonic stem cell enhancers.

Lex Fridman (48:53.280)

So the genetic variants predisposing you to be taller or shorter are in fact acting in

Lex Fridman (48:57.480)

developmental stem cells, makes complete sense.

Manolis Kellis (49:01.380)

If you looked at inflammatory bowel disease, you basically found inflammatory, which is

Lex Fridman (49:05.920)

immune, and also bowel disease, which is digestive.

Lex Fridman (49:09.880)

And indeed we saw a double enrichment both in the immune cells and in the digestive cells.

Lex Fridman (49:15.800)

So that basically told us that this is acting in both components.

Manolis Kellis (49:19.040)

There's an immune component to inflammatory bowel disease and there's a digestive component.

Lex Fridman (49:23.460)

And the big surprise was for Alzheimer's.

Manolis Kellis (49:25.960)

We had seven different brain samples.

Manolis Kellis (49:29.160)

We found zero enrichment in the brain samples for genetic variants associated with Alzheimer's.

Lex Fridman (49:36.240)

And this is mind boggling.

Lex Fridman (49:38.020)

Our brains were literally hurting.

Lex Fridman (49:40.040)

What is going on?

Lex Fridman (49:42.000)

And what is going on is that the brain samples are primarily neurons, oligodendrocytes, and

Manolis Kellis (49:49.080)

astrocytes in terms of the cell types that make them up.

Lex Fridman (49:54.120)

So that basically indicated that genetic variants associated with Alzheimer's were probably

Manolis Kellis (49:59.400)

not acting in oligodendrocytes, astrocytes, or neurons.

Lex Fridman (50:04.560)

So what could they be acting in?

Manolis Kellis (50:05.960)

Well, the fourth major cell type is actually microglia.

Lex Fridman (50:10.200)

Microglia are resident immune cells in your brain.

Manolis Kellis (50:13.720)

Oh, nice.

Lex Fridman (50:15.880)

They immune.

Manolis Kellis (50:16.880)

Oh, wow.

Lex Fridman (50:17.880)

They are CD14 plus, which is this sort of cell surface markers of those cells.

Lex Fridman (50:24.160)

So they're CD14 plus cells, just like macrophages that are circulating in your blood.

Lex Fridman (50:30.200)

The microglia are resident monocytes that are basically sitting in your brain.

Manolis Kellis (50:35.640)

They're tissue specific monocytes.

Lex Fridman (50:38.400)

And every one of your tissues, like your fat, for example, has a lot of macrophages that

Manolis Kellis (50:42.840)

are resident.

Lex Fridman (50:43.920)

And the M1 versus M2 macrophage ratio has a huge role to play in obesity.

Lex Fridman (50:49.560)

And so basically, again, these immune cells are everywhere, but basically what we found

Manolis Kellis (50:53.440)

through this completely unbiased view of what are the tissues that likely underlie different

Manolis Kellis (50:59.080)

disorders, we found that Alzheimer's was humongously enriched in microglia, but not at all in the

Lex Fridman (51:08.080)

other cell types.

Lex Fridman (51:09.080)

So what are we supposed to make that if you look at the tissues involved, is that simply

Lex Fridman (51:15.480)

useful for indication of propensity for disease, or does it give us somehow a pathway of treatment?

Manolis Kellis (51:24.640)

It's very much the second.

Lex Fridman (51:26.120)

If you look at the way to therapeutics, you have to start somewhere.

Lex Fridman (51:33.900)

What are you going to do?

Manolis Kellis (51:34.900)

You're going to basically make assays that manipulate those genes and those pathways

Manolis Kellis (51:42.040)

in those cell types.

Lex Fridman (51:43.780)

So before we know the tissue of action, we don't even know where to start.

Manolis Kellis (51:49.200)

We basically are at a loss.

Lex Fridman (51:51.040)

But if you know the tissue of action, and even better, if you know the pathway of action,

Manolis Kellis (51:54.720)

then you can basically screen your small molecules, not for the gene, you can screen them directly

Lex Fridman (52:00.280)

for the pathway in that cell type.

Lex Fridman (52:02.640)

So you can basically develop a high throughput multiplexed robotic system for testing the

Manolis Kellis (52:10.480)

impact of your favorite molecules that you know are safe, efficacious, and sort of hit

Manolis Kellis (52:16.240)

that particular gene and so on and so forth.

Manolis Kellis (52:18.940)

You can basically screen those molecules against either a set of genes that act in that pathway

Manolis Kellis (52:25.720)

or on the pathway directly by having a cellular assay.

Lex Fridman (52:29.820)

And then you can basically go into mice and do experiments and basically sort of figure

Manolis Kellis (52:33.400)

out ways to manipulate these processes that allow you to then go back to humans and do

Manolis Kellis (52:38.800)

a clinical trial that basically says, okay, I was able indeed to reverse these processes

Manolis Kellis (52:43.200)

in mice.

Lex Fridman (52:44.200)

Can I do the same thing in humans?

Lex Fridman (52:46.240)

So the knowledge of the tissues gives you the pathway to treatment, but that's not the

Lex Fridman (52:51.820)

only part.

Manolis Kellis (52:52.820)

There are many additional steps to figuring out the mechanism of disease.

Lex Fridman (52:57.280)

So that's really promising.

Manolis Kellis (52:59.080)

Maybe to take a small step back, you've mentioned all these puzzles that were figured out with

Manolis Kellis (53:04.360)

the Nature paper for, I mean, you've mentioned a ton of diseases from obesity to Alzheimer's,

Manolis Kellis (53:13.960)

even schizophrenia, I think you mentioned.

Lex Fridman (53:17.720)

What is the actual methodology of figuring this out?

Lex Fridman (53:20.720)

So indeed, I mentioned a lot of diseases and my lab works on a lot of different disorders.

Lex Fridman (53:26.040)

And the reason for that is that if you look at biology, it used to be zoology departments

Lex Fridman (53:39.500)

and botanology departments and virology departments and so on and so forth.

Lex Fridman (53:43.680)

And MIT was one of the first schools to basically create a biology department, like, oh, we're

Manolis Kellis (53:47.680)

going to study all of life suddenly.

Lex Fridman (53:49.640)

Why was that even a case?

Manolis Kellis (53:51.720)

Because the advent of DNA and the genome and the central dogma of DNA makes RNA makes protein

Lex Fridman (53:58.480)

in many ways, unified biology.

Manolis Kellis (54:01.600)

You could suddenly study the process of transcription in viruses or in bacteria and have a huge

Manolis Kellis (54:07.480)

impact on yeast and fly and maybe even mammals because of this realization of these common

Manolis Kellis (54:15.040)

underlying processes.

Lex Fridman (54:17.760)

And in the same way that DNA unified biology, genetics is unifying disease studies.

Lex Fridman (54:27.180)

So you used to have, I don't know, cardiovascular disease department and neurological disease

Manolis Kellis (54:39.440)

department and neurodegeneration department and basically immune and cancer and so on

Lex Fridman (54:47.640)

and so forth.

Lex Fridman (54:48.640)

And all of these were studied in different labs because it made sense, because basically

Manolis Kellis (54:53.600)

the first step was understanding how the tissue functions and we kind of knew the tissues

Lex Fridman (54:57.560)

involved in cardiovascular disease and so on and so forth.

Lex Fridman (55:00.680)

But what's happening with human genetics is that all of these walls and edifices that

Lex Fridman (55:05.760)

we had built are crumbling.

Lex Fridman (55:08.480)

And the reason for that is that genetics is in many ways revealing unexpected connections.

Lex Fridman (55:16.560)

So suddenly we now have to bring the immunologists to work on Alzheimer's.

Manolis Kellis (55:21.480)

They were never in the room.

Lex Fridman (55:22.680)

They were in another building altogether.

Manolis Kellis (55:25.920)

The same way for schizophrenia, we now have to sort of worry about all these interconnected

Lex Fridman (55:31.600)

aspects.

Manolis Kellis (55:33.200)

For metabolic disorders, we're finding contributions from brain.

Lex Fridman (55:37.500)

So suddenly we have to call the neurologist from the other building and so on and so forth.

Lex Fridman (55:41.340)

So in my view, it makes no sense anymore to basically say, oh, I'm a geneticist studying

Lex Fridman (55:49.200)

immune disorders.

Manolis Kellis (55:50.200)

I mean, that's ridiculous because, I mean, of course in many ways you still need to sort

Lex Fridman (55:55.360)

of focus.

Lex Fridman (55:56.480)

But what we're doing is that we're basically saying we'll go wherever the genetics takes

Lex Fridman (56:01.080)

us.

Lex Fridman (56:02.440)

And by building these massive resources, by working on our latest map is now 833 tissues,

Manolis Kellis (56:10.440)

sort of the next generation of the epigenomics roadmap, which we're now called epimap, is

Manolis Kellis (56:15.560)

833 different tissues.

Lex Fridman (56:18.120)

And using those, we've basically found enrichments in 540 different disorders.

Manolis Kellis (56:24.620)

Those enrichments are not like, oh great, you guys work on that and we'll work on this.

Lex Fridman (56:29.340)

They're intertwined amazingly.

Lex Fridman (56:31.980)

So of course there's a lot of modularity, but there's these enhancers that are sort

Lex Fridman (56:36.120)

of broadly active and these disorders that are broadly active.

Lex Fridman (56:39.040)

So basically some enhancers are active in all tissues and some disorders are enriching

Lex Fridman (56:43.480)

in all tissues.

Lex Fridman (56:44.480)

So basically there's these multifactorial and this other class, which I like to call

Lex Fridman (56:49.160)

polyfactorial diseases, which are basically lighting up everywhere.

Lex Fridman (56:54.560)

And in many ways it's, you know, sort of cutting across these walls that were previously built

Lex Fridman (57:00.000)

across these departments.

Lex Fridman (57:01.760)

And the polyfactorial ones were probably the previous structural departments wasn't equipped

Lex Fridman (57:07.040)

to deal with those.

Manolis Kellis (57:08.040)

I mean, again, maybe it's a romanticized question, but you know, there's in physics, there's

Lex Fridman (57:14.680)

a theory of everything.

Lex Fridman (57:16.920)

Do you think it's possible to move towards an almost theory of everything of disease

Lex Fridman (57:22.400)

from a genetic perspective?

Lex Fridman (57:24.100)

So if this unification continues, is it possible that, like, do you think in those terms, like

Lex Fridman (57:29.640)

trying to arrive at a fundamental understanding of how disease emerges, period?

Manolis Kellis (57:35.720)

That unification is not just foreseeable, it's inevitable.

Lex Fridman (57:41.680)

I see it as inevitable.

Manolis Kellis (57:43.600)

We have to go there.

Lex Fridman (57:45.240)

You cannot be a specialist anymore.

Manolis Kellis (57:48.340)

If you're a genomicist, you have to be a specialist in every single disorder.

Lex Fridman (57:53.840)

And the reason for that is that the fundamental understanding of the circuitry of the human

Manolis Kellis (57:59.960)

genome that you need to solve schizophrenia, that fundamental circuitry is hugely important

Lex Fridman (58:07.960)

to solve Alzheimer's.

Lex Fridman (58:09.600)

And that same circuitry is hugely important to solve metabolic disorders.

Lex Fridman (58:13.100)

And that same exact circuitry is hugely important for solving immune disorders and cancer and,

Manolis Kellis (58:20.040)

you know, every single disease.

Lex Fridman (58:22.260)

So all of them have the same sub task.

Lex Fridman (58:26.680)

And I teach dynamic programming in my class.

Lex Fridman (58:29.880)

Dynamic programming is all about sort of not redoing the work.

Manolis Kellis (58:34.400)

It's reusing the work that you do once.

Lex Fridman (58:37.280)

So basically for us to say, oh, great, you know, you guys in the immune building go solve

Manolis Kellis (58:42.240)

the fundamental circuitry of everything.

Lex Fridman (58:44.240)

And then you guys in the schizophrenia building go solve the fundamental circuitry of everything

Manolis Kellis (58:47.680)

separately, is crazy.

Lex Fridman (58:50.080)

So what we need to do is come together and sort of have a circuitry group, the circuitry

Manolis Kellis (58:56.080)

building that sort of tries to solve the circuitry of everything.

Lex Fridman (58:59.520)

And then the immune folks who will apply this knowledge to all of the disorders that are

Manolis Kellis (59:05.920)

associated with immune dysfunction and the schizophrenia folks will basically interacting

Lex Fridman (59:12.460)

with both the immune folks and with the neuronal folks.

Lex Fridman (59:15.560)

And all of them will be interacting with the circuitry folks and so on and so forth.

Lex Fridman (59:19.000)

So that's sort of the current structure of my group, if you wish.

Lex Fridman (59:22.320)

So basically what we're doing is focusing on the fundamental circuitry.

Lex Fridman (59:27.200)

But at the same time, we're the users of our own tools by collaborating with many other

Manolis Kellis (59:34.020)

labs in every one of these disorders that we mentioned.

Manolis Kellis (59:37.480)

We basically have a heart focus on cardiovascular disease, coronary artery disease, heart failure

Lex Fridman (59:42.880)

and so on and so forth.

Lex Fridman (59:44.280)

We have an immune focus on several immune disorders.

Manolis Kellis (59:48.900)

We have a cancer focus on metastatic melanoma and immunotherapy response.

Manolis Kellis (59:55.580)

We have a psychiatric disease focus on schizophrenia, autism, PTSD, and other psychiatric disorders.