Steve Hsu: Welcome to Manifold. My guest today is Abdel Abdellaoui, a well-known researcher in genetics and behavior genetics. He is a professor at the University of Amsterdam. I had the pleasure of dinner with Abdel in Utrecht, which is a city near Amsterdam. I was in Holland for the first time, actually in my life last summer to give two talks, one both in physics, one at Utrecht, and one at the University of Amsterdam.

And Abdel was kind enough to come over to Utrecht, too, but because of the way the schedules worked out, he had to come to Utrecht. And so, we met and had a lovely, I think, Vietnamese dinner out in the street on a beautiful day in the summertime. And you can't imagine how. Lovely. Holland is in the summer, so I'm really glad I was able to get Abdel to join the podcast.

Welcome, Abdel.

Abdel Abdellaoui: Thank you for having me. Steve, so I have to correct you. I'm actually not a Professor. I'm, I've just been, promoted to assistant professor, but I was a, a postdoc before, then I, yeah,

Steve Hsu: Okay. I apologize. I should have said a professor maybe instead of associate professor.

Abdel Abdellaoui: Yeah.

Steve Hsu: A professor is good enough, right?Well, since you are very precise and maybe too self-deprecating, let me add something positive for our listeners. Your work has been cited thousands of times, and the reason I reached out to you was because I had read many of your papers and I knew I was going to be in the area of Amsterdam last summer, so I wanted to meet with you, in person if possible. So, assistant, or full professor, whichever it is, you're, you're very accomplished in your field.

Abdel Abdellaoui: Yeah. Thank you. Yeah, yeah, yeah. I knew about you and your activities, of course. So, I was happy to join you at dinner and I enjoyed our conversation very much. I like that you're actually a physics professor. I'm always very fascinated by physics, but I don't know too much about it.

Steve Hsu: Well, we can certainly, we can certainly discuss it, but I think what the audience wants to hear is, about your work in genomics. Let's start with that. So, you grew up in Holland and I'm curious a, what it was like, because I think you're from an immigrant background just like I am curious about what it was like growing up as an immigrant in Holland, how you got interested in genomics, and anything you want to say about your early life.

Abdel Abdellaoui: Yeah, sure. So I was, indeed, born 40 years ago in Amsterdam from my immigrant parents. My parents are from Morocco. They came here in the late sixties, early seventies. They come from a rural area in Morocco, they didn't have much education and many opportunities there. So, they came to work here.

Abdel Abdellaoui: They had; they had a pretty rough life. My dad started working in the mines in France and then in the factories, an iron factory in the Netherlands. And my mom was just a stay-at-home mom. We did not speak the language, and we just gave all her energy and love to the kids. So, we, we, and, and made sure that we had everything we needed.

So, I actually had a pretty good childhood. I grew up in a neighborhood that was also full of migrants and locals. And, it was, like when I look back at it now, like there were a lot of drug dealers on the corners but, When I was a kid, I wasn't aware of all of that.

And I just remember a lot of joyfulness and playing, and I had everything I needed. Yeah, so that's how I remember my childhood. and it was also in a time where there was a lot of investments in, in, those kinds of neighborhoods. So, we had a lot of facilities for kids to play.

We had something called a build house in, in Dutch that the literal translation is Neighborhood House, where a lot of kids from the neighborhood could come and play video games or participate in sporting tournaments, their organized trips. So, I felt like I got taken care of pretty well. so. When I finished high school, I went to study informatics at the university here, Amsterdam, Free University, because I didn't really know what I wanted to do with my life, but I knew that there were a lot of job opportunities there and I was pretty good at working with computers. But then after two years, I realized that I found it, like I was, I could do it, but it was just a little too boring for me to learn so many details, so many take books about how computers work. So, I decided to change topics and studied psychology. I was more interested in humans and also, A bit, searching, of who, who I was and what was going on in this social environment. So, I thought I would enjoy psychology.

And I also found it kind of exciting choosing something without really knowing what I would end up doing with it later. So just following my interests. so that was very exciting to me. Changed a lot about how I looked at the world. I was raised quite religiously. For example, I believed in a soul and in heaven and hell.

Abdel Abdellaoui: And then during that education, I got in touch with things like philosophy and, how, how scientific, the history of scientific research and critical thinking. So, I made a big shift in the way I looked at the world and I had to sort of learn how to create a new world for myself and, and, and use science at, at the center of that.

There was one specific course called biological psychology, which was just about the brain. And that made me realize that everything really is happening in that organ in, in, our head. So, I chose. To specialize in a neuroscience master. And within that, masters, there was a genetics course. And that course got me fascinated with our DNA.

I realized this, that molecule contained the answers of, of where we came from and why we are the way we are today. So, I started working at the department that gave that course. Before I finished my masters, I was already working there and helping them, contacts, and, and locating twins all over the countries.

They did a lot of twin studies, which is how the department started to do their genetics research in the eighties. And then when I finished my master's degree, I was so embedded in the department that they offered me a PhD position. And my PhD was on a combination of behavioral genetics and population genetics.

So, my PhD was supposed to be about the genetics of depression, so I was actually being paid to do that. but when I got my hand on their first genotype dataset, this was around 2009, 2010 when Genome-Wide Association Studies were just starting, to, to become big enough to be successful.

I can explain what a genome-wide association study is. Maybe, probably later. But I, I, I, I got my hands on the DNA of thousands of people and I was just cleaning the data and. looking at the big patterns of variation before I, could, could link it to something as complex as depression. I was just looking at what, what does the data look like?

And then I removed some of those quality differences that were captured by the big patterns of variation. And then recompute it, what, what the, do the big patterns of variation look like now and then, there were some interesting patterns, but I wasn't sure what they were. And then I plotted them on a map, and I saw very beautiful patterns that separated like the north of the country from the south of the country.

And then I realized that they represent different ancestries. So, if you have like, half a million or, or a million genetic variants, and you do what we call a principal component analysis on them, it summarizes the biggest patterns of variation. And those reflect ancestry and correlate beautifully with geography because the closer you live to someone, to more ancestors, you share, share with that person.

So then more than half of my PhD thesis ended up being about these ancestry differences and putting them in a historical context and seeing how they were influenced by things like migration and partner choice. And these are all very important processes to understand if you want to link those genetic variants, eventually.

Complex traits like, major depression. So that was my PhD thesis. The title was Behavior Genetics With an Arrow in between. That points both ways because I started being interested in what genes influence behavior, but I ended up writing about how our behavior, things like migration and partner choice, influence the genetic makeup of a population and how you have to, account for that in, the studies that are trying to find genes that influence complex traits or these are genome-wide, association studies.

So, I hope I'm not all over the place.

Steve Hsu: No, that's, that's perfect.

And, I think, I think most of the listeners to this podcast kind of know what a GWAS is.

They know about statistical associations between a particular genetic variant and say increased risk or decreased risk of a particular disease condition. I think people understand that.

You mentioned something really interesting, which is that at the very earliest stages of your looking at the, I guess this was Dutch data, you could immediately see some kind of north, south gradient in ancestry. And that was even before you were looking at the particular trait like major depression.

Abdel Abdellaoui: Yeah, yeah, it's true. Yeah. Yeah. So, we, there, there are like three major rivers that are going, through the Netherlands that separate the north of the south. So, you have a geographic barrier and then you have a cultural barrier. The north of the river has been Protestant for hundreds of years, and the south has been Catholic for hundreds of years.

Those people have been, have had their own universities, their own political parties, and they've been only making babies with each other for many centuries. And that's very, well detectable in, in if, in in the genetic data. So, we had DNA from people literally born all over the Netherlands and the maps were striking.

And then, and then you have a pocket. on the west side of the country where all of the major cities are. And those cities all were sort of a mixture of those northern and southern, ancestry, which by the way look a lot like Northern European versus Southern European, ancestry, like the gene that showed the strongest difference between the north and the south was the gene that codes for whether we have blue or brown, brown eyes. the, the. Difference between north and south was so big that it's almost certain that there has been some kind of evolutionary selection pressure on this variant. But another reason that the difference is so big is because of the huge effect that it has. If you have two copies of that variant, then your chances of having blue eyes are like 99%.

Abdel Abdellaoui: So, if there is an evolutionary selection pressure, it focuses, almost, exclusively on that variance. So, yeah.

Steve Hsu: Sorry to interrupt you. I hate to divert us from, you know, one of the interesting aspects of your research is that you, you work on things which are, related to psychology and behavioral traits and things like this. But now you're mentioning a very, something that, you know, people under a lot of people understand is being genetically determined, which is eye color.

Abdel Abdellaoui: Hmm.

Steve Hsu: I'm curious, so you just said something which is pretty interesting. You said you're, you, you have a strong suspicion that there really was differential selection on this, so

Perhaps in the north of Holland there was more selection for having blue eyes than in the South. Is that [right]?

Abdel Abdellaoui: Yes, but it's probably more Northern Europe. I'm not sure if this whole selection thing took place in the Netherlands. There are multiple reasons why people would have, like the, those three major rivers were also a border for the Roman Empire, so there was a lot of migration to, from the south to the, to the Netherlands until that border.

So what was your question exactly about this?

Steve Hsu: No, I think you kind of answered it. I guess I misunderstood what you were saying. So, I think you're saying generally in Europe there was some kind of selection pressure that led to higher prevalence of blue eyes in the North. And the reason it's reflected in Holland particularly maybe has more to do with migration from the south of Europe, you know, up to the river boundary, and then from the north of Europe down to the river boundary.

Abdel Abdellaoui: I think so. Yeah.

Steve Hsu: Yeah.

It's quite interesting. Let me not, let me not digress too much on eye color because I think the most interesting part of your work has to do with some of the phenotypes that you've studied. And let me just list them for the audience. So, you've studied things like major depression, suicide attempt,psychiatric disorders like schizophrenia, major depressive disorder, and bipolar disorder.

Steve Hsu: And you've also studied educational attainment,

Abdel Abdellaoui: Yeah.

And I, I think you explained already because you kind of came to this from the psychology department, it was natural for you to study conditions like that. Maybe you could just say a few words about what is special about those kinds of conditions versus things that other medical researchers might study, like heart attack or diabetes or something.

Abdel Abdellaoui: Yeah. Well, they are more, they are more complicated, that's for sure to study, genetically than, because they're under less direct biological control, if you will. There's a lot more interaction with, environment, social environment in particular that influences those traits. But that may, Yeah, that makes, for me, that makes them more interesting because you can incorporate, like our, our behavior and how we also, influence.

which environment we are exposed to. So, so if, if you do a GWAS on something like, major depression, So in a GWAS we, we test these, millions of variants, which each have very tiny effects. And there is, there is not a gene that just makes you more depressed. All these genes are influencing many biological processes and many components that influence different outcomes.

And those outcomes influence what environment you are exposed to, but they are also influenced by the environment. So, there is a lot that is being captured in that GWAS signal. yeah. And so, I just find it exciting that we can measure something like DNA where we know that that's like where it starts and it's, a molecule that we've never even seen with our bare eyes.

So, it feels almost like magic that we can measure those nucleotides and then connect them to such a faraway complex outcome. and it works. And the associations are, they replicate and you can use them to predict the risk in individuals that are not in the data set where you detected that association.

So, we know that the connection is real, but there's so much in between those genetic variants and that complex outcome that just gives us a lot to work on. And I get excited about that because I'm looking for work to do.

Steve Hsu: That's great. So, so I, I, I've looked a little bit at schizophrenia, for example, and my impression now is that there are some fairly strong polygenic predictors for schizophrenia and for the audience. What that means is that if you, if you have someone's DNA and you compute their risk score for schizophrenia, people that are in the top, I'm, I'm just making these numbers up.

I don't have them in front of me, but, you know, people who are in the top few percent of risk, genetic risk for schizophrenia as predicted by the score, they might have, I don't know, multiple, five times more risk, absolute risk of eventually having schizophrenia later in their life. I think that's my general impression. Maybe you could say a little bit about what the current status is for major depression.

Abdel Abdellaoui: Oh, it's not as good as schizophrenia. That's also because it's a less heritable trait. The heritability of depression is around 30, 35% as estimated with twins. And then for schizophrenia, it's like 80, 80%. So, for schizophrenia, they needed it, they were much quicker with finding significantly associated genes and polygenic scores that were predictive.

But major depression scores are getting better. But the thing with depression, like what it also says, with how heritable something. It also has to do with, like heritability means the number of individual differences that are explained by genetic differences. Which means that if, if we all would live in the exact same environment, then major depression should be a hundred percent heritable, theoretically.

If we all had the same environmental circumstances, then the individual differences that remain would have to be explained by genetics. So, it’s, relative measure, but the fact that though we have more trouble detecting depression genes than schizophrenia genes means that there's a lot more variation in the environment that makes people, some people more depressed than others.

Steve Hsu: But in the, in the case of, I think you said the estimated heritability from twin studies for schizophrenia is very high. Did you say 80%?

Abdel Abdellaoui: 80, 80%. Yeah. From twin studies.

Steve Hsu: Yeah. That's very high. So, and, and, and we know that it runs in families, the condition?

Abdel Abdellaoui: Yeah.
Steve Hsu: That's much less so I guess for depression, you're saying?

Abdel Abdellaoui: Yeah. I think like for depression, I find that studying the genetics of depression interesting and worthwhile because it can teach us something about the underlying both biological and social mechanisms in scientific research but from a clinical perspective.

And that's what I, I think you were hinting at when you were asking about polygenic scores, how well they predict. Usually, when people talk about that, they also are hoping that they will someday predict well enough to be able to use in the clinic. But for something like depression, I, I, I think it would be, there's a lot more to be won by looking at the environment that people living in. yeah,

Steve Hsu: They’re, though, crude-like estimators. Like, someone's social economic status or something that, that does a good job [unclear] elevated risk for depression?

Abdel Abdellaoui: Yeah, probably. So, I'm not sure how far we are with the predictability of the polygenic score for depression. But like the, under the current social circumstances, the highest we can get is probably around 30%. And that's if you also include rare variants. Without a rare variance, we probably get to 15% or something. And so, I'm not exactly sure about the predictive ability of other non-genetic factors for the risk of major depression. So, I'm not the clinician, so I've, I, and I never see patients, so I'm not that knowledgeable about what is and is not useful in the clinic. especially for something like depression.

I think genetics could be more useful in the near future for things like schizophrenia and bipolar disorder, but yeah, depression is just, yeah, it happens to be the least hurtful of all the psychiatric disorders. That's why maybe I'm sounding so hesitant.

Steve Hsu: Yeah, I maybe should not have focused the discussion on major depression. I just, I just noticed that I think maybe your most cited paper is actually on major depression, and I know very little about it, so I just thought he would ask you a little bit about it.

Let's come back to the idea of gene-environment interactions or correlations across geographical regions. I think that's something that you've worked on. Maybe you could give us an example of that kind of result.

Abdel Abdellaoui: Yeah. Yeah. Yeah. So that's a topic I'm, I feel more qualified to talk about than the major depression. So, it's my most cited paper, but I'm just a co-author on that paper,

Steve Hsu: No, I understand.

Abdel Abdellaoui: I did some analysis.

Steve Hsu: Yeah. I probably should not have pushed in that direction, but I just didn't know much about it myself, so I was curious.

Abdel Abdellaoui: But the stuff with the gene-environment, correlation, and the geography. So those are the studies that I've actually, led and, and did entirely abroad, or not did entirely, but that I led. And a lot of that also came about by combining genetic data with geography like I did in the Netherlands.

So, in the Netherlands, I looked at the relationship between those patterns that reflect ancestry differences and geography. But then, around 2018, UK Biobank was released, which is, half million individuals genotyped in, that were born all over Great Britain, which is, has, has been. Hugely helpful for genetics, for, a lot of other subfields of bi biology probably, but especially for genetics.

Abdel Abdellaoui: It has been crucial in the, in the progress that we have been able to, to make if there was, it's too bad that they don't give Nobel prizes for institutes because the UK Biobank, the impact that had on genetics research was incredible. It gave me the opportunity to combine geography, not just with those ancestry differences, which are the biggest patterns. So, for those effects, I had enough of, for analyzing the six, 7,000 Dutch individuals that I have. But UK Biobank allowed me to look at these polygenic scores and, Plot those on a map and look at the geographic distribution of those. So that's an article that came out in 2019.

I made a polygenic course for more than 30 traits, so we have had more than 30, for, physical and mental health outcomes. Things like personality height, BMI. And so, these were all gas that did not include UK Biobank, where you have the genetic effect estimates from those GWAS, and you used those to build a polygenic score in the UK.

Abdel Abdellaoui: In UK Biobank, that is predicted for more than these 30 traits. We know where these people in UK Biobank were born. We know where they live now. And when I looked at whether these polygenic scores were clustered geographically, or whether they showed regional differences, we saw that nearly all of the polygenic scores showed significant geographic clustering.

So, there were significant differences between regions, but when you plot them on a map, all those geographic distributions looked very similar to those of the ancestry differences, because I also plotted those ancestry differences for the UK Biobank. They largely separate Wales from England and Scotland. So, do you see the cultural barriers if you will?

But then when we controlled for those ancestry differences, we saw that the geographic cluster dropped for almost all of the traits except for educational attainment. So, the polygenic score that predicted how long you went to school, those showed very significant regional differences. And when you plot those on a map, then those distributions look exactly the same as the distribution of social economic status.

So, of how well a neighborhood is doing economically. And you could very, like the biggest, most, clearest patterns were the, like the coal mining regions of Great Britain versus the rest of the country. So, in the coal mining regions, the coal mining industry collapsed in the last century between the thirties and the eighties. And then the joblessness increased and a lot of economic struggles and those became the poorest neighborhoods of the country. And that was visible when you looked at the distribution of this polygenic score for educational attainment. And what we also saw was that it was probably driven by migration, at least increasing by, migration because the people that were born in the coal mining regions and migrated away, those had a higher polygenic score for educational attainment than the entire country on average.

So, there was sort of a brain drain going on, but on a genetic level. And you saw those differences increase, throughout the last century. So, then we get to the gin environment correlation because these, those, these genes, so they cluster in these regions. But these regions also show huge differences environmentally. So, these polygenic scores are all associated with whether you live in a poor neighborhood or not, and what, whether you live in a poor neighborhood or not, influences a lot of your physical and mental health outcomes, including educational attainment itself. But also, things like obesity, like the number of fast-food restaurants. If you make a map of the number of fast-food restaurants, you get exactly the same map as the map of the polygenic scores.

So, we published an article about the geographic distribution in 2019. And then just, there was a natural human behavior. And then just a month ago in [unclear] Genetics, we published the article about the actual gene environment correlation. So, one of the things we showed there was we looked at sibling pairs, and the sibling with the higher polygenic score, if you will, was more likely to migrate to better, neighborhood.

So, these gene-environment correlations, they happen on a geographic level and also on a family level, right? So, if you have two siblings in a family, they inherited those genes for, let's say educational attainment from their parents, but they also inherit the environment from their parents.

So, the genes in those siblings will, will, correlate with the environment in the parents. So, if you take a polygenic score and then, then it predicts educational attainment between families a lot better than within families. So, it, predicts the difference between siblings, much worse than the difference between two people from two different families because the polygenic score predicts the genetic effects as well as the environmental effects on a family level.

Abdel Abdellaoui: So, we saw that the same thing happens on a geographic level as well. The polygenic score was, was predicting, it, it not just educational attainment, but the whole wide, we looked at this for more than 50 traits. And, and for a lot of traits, those polygenic scores captured environmental effects on both a family level and a geographic level.

So that the paper consisted of two parts. One part where we looked at siblings on a family level and geographic level, how well the polygenic course captured those environmental effects. And then in the second part of the paper, we just did a genome-wide association study in, a couple of hundred thousand UK Biobank individuals on more than 50 traits.

Abdel Abdellaoui: And then we looked at what happens with this, these GWAS results if you control for the geographic region where people live in. And what happened? And it was, this was for a wide range of traits for things like obesity, for depression, for substance use, for risk taking behavior. For blood pressure, we looked at 56 traits in total.

If you control for geographic regions, then you remove genetic effects that are associated with educational attainment from the signal. Because if you don't control for geographic regions, you also capture those effects of the genes on educational attainment that determine in which neighborhood you end up living.

So, if you have a genetic makeup that for some reason makes you do better in school, you will be put into a better school from a very young age. You will grow up to get a better job. You will be able to afford a house in a better neighborhood, and you will live in more healthy circumstances that will have a positive influence on things like your blood pressure, your mental health, your height, even, And that will make those genes associated with all of those physical and mental health outcomes in a genome-wide association study.

So that makes these genetic studies a lot more complicated to interpret than we previously assumed. Probably

Steve Hsu: Wow, you. Okay. So, you've, you've covered a lot there. In fact, I think you just summarized two landmark papers, right? One, one from 2019 and then the other one just came out relatively recently. Do you mind if I just backtrack a little bit for the audience?

Abdel Abdellaoui: Yeah.

Steve Hsu: Okay, so in the earlier paper, one of the things you did was you first noticed geographical variation in the polygenic scores themselves.

Right? So across, depending on what part of the UK the person is from, there's some, at least, you know, maybe it's small, but it's statistically detectable difference in, say, the polygenic EA score, educational attainment score.

Abdel Abdellaoui: Yeah,

Steve Hsu: And, but then like in, in the later work, you tried to control for that so that, you, you could, you could remove the, you know, the influence of the, the geographical origin from that particular score.

Abdel Abdellaoui: Yeah. Of the environmental effects that correlate with those, genes on a geographic level.

Steve Hsu: Right.

And already in the earlier paper, you detected migration, which was correlated to the educational attainment, polygenic score so that people from a depressed region were more likely to migrate away, say perhaps to a big city or more favorable region if they had better than average polygenic EA scores. That was one of the main results of the first paper.

Abdel Abdellaoui: Yes, that's correct.

Yeah.

Steve Hsu: And, you know, maybe not surprising in a sense because, there's always this narrative of like, whether it's in, you know, the urbanization of China in the late 20th century or of America in the late 19th century or something, there's this idea that, you know, there are lots of people in agriculture and they're not making much money.

And maybe the more energetic or the bright ones end up going to the big city and succeeding there. And so maybe you're seeing the genetic imprint, the genetic, version of that story. Is that fair?

Abdel Abdellaoui: Yeah. Yeah. That was exactly how my mom reacted when I explained it to her. She said, Oh, yeah, I see the same stuff happening in Morocco in the small villages.

Steve Hsu: Yeah, I think it's a familiar story, about urbanization or migration.

Now in the later paper, one of the things I think probably, I'm not going to be able to summarize everything you said, because you said a lot for the audience, but, but one of the things you said was that you're trying to disentangle the multiple impact of the educational attainment polygenic score on other things.

So, for example, if you do well in school, you end up with a better career, you end up living in a better neighborhood and you end up eating better food and, maybe less stress from day to day. And that has additional impacts on your health risks across various categories or your BMI or even your height. And so now you're able to actually see some of those effects?

Abdel Abdellaoui: Yes, that is correct. Yeah. We are able to see some of those effects. We are able to control for them to some extent if in a genome MI association study, because if you do a, a, a GWAS on something like schizophrenia or blood pressure, you, you, you're not interested in. Well, I think most people are not interested in capturing genes that influence education.

So, we try to minimize them, but it's very difficult to eliminate them all together because the geographic location is just a, a crude proxy for the history of your environment and your social environment for if to, in order to control for it completely, we would have to know exactly what the social environment was of the individuals throughout their lives.

And that's just, that's, that's going to be hard, I think at the nitty-gritty level for the researcher, I just maybe want to describe it this way. You're doing a kind of statistical analysis of genetic variants that might influence BMI, body mass index, or, you know, how, how, how, how much fat you have, basically. and you discover some variant that is associated with higher BMI.

Abdel Abdellaoui: Yeah.

Steve Hsu: And the effect you're pointing out is that, well, this particular genetic variant might not be directly affecting the metabolism of fat deposition or that regulates your hunger instincts, feelings or whatever. It might actually be some variant that's affecting your ability to do well in school.

And it's through the eventual consequence that the people who do well in school end up in environments where the food is healthier. And so, it ends up being picked up by statistical analysis or machine learning, as being a BMI variant. But then, the causal pathway is really complicated.

Abdel Abdellaoui: Yeah. Yeah. Because there is also not a genetic variant that just influences, that only influences how well you do in school. Like that. GWAS on educational attainment is also a composite of many different underlying traits. We also did one study where we subtracted the genetic effects on intelligence from the GWAS signal, and that was on, that was about half of the signal on of educational attainment was intelligence.

The other half was stuff related to personality and physical and mental health. And, and, and, and all of those separate components are probably also super complicated, composition of, of multiple trades. So yeah,

Steve Hsu: Just to repeat what, Just to repeat what you just said, for the audience, educational attainment, which is based on a phenotype of really just how many years of schooling were you able to complete

in your life? I think most people would find it plausible that Okay, if, If there's something that's affecting how well your brain functions, which we might call intelligence or something, that would certainly have an impact on years of education because if you're good at school, you're likely to get more school.

But then there were other things like, well, how healthy are you? You know, if you're not healthy, maybe you can't sustain, you know, you can't survive through your college program and you drop out. Or if you're not conscientious or ambitious for learning or something, which is not directly just intelligence, but some suitability for schooling, that could also be another genetic factor, which is independent from intelligence.

Abdel Abdellaoui: Yeah, yeah. Exactly.

Steve Hsu: Yeah.

So, and you're actually in the process of kind of teasing those things apart.

Abdel Abdellaoui: Yes, those, That's what we're trying to do. Yeah. Yeah. Step by step, we're starting to realize how complicated these signals are, and that's what my research is probably going to be focused on in the coming years. And what probably the whole field should be focusing on is, trying to disentangle all these different components that make up the GWAS signal.

So that's one thing we should do. The other thing is we should keep increasing the GWAS sample sizes, because the bigger they get, the better the genetic effect estimates are. So, the better we can then distinguish between all these different components, and we should also start expanding data collection to other populations that live in different social and environmental circumstances because the, the effects that we are capturing now are almost all from, countries, with, with, where people live with the European ancestry, Western countries, most of them probably from Great Britain, from UK Biobank.

Steve Hsu: On that last point. On that last point, I don't know if you saw the poster that my group presented at ASHG, the American Society for Human Genetics, which just happened, I guess it was last week. I wasn't there. But

Abdel Abdellaoui: I wasn't there either. Yeah.

Steve Hsu: one guy from our group went and presented this poster, and it was based on our collaboration with the Taiwan Precision Medicine Initiative, and they now have genotyped half a million.

Abdel Abdellaoui: Oh, wow.

Steve Hsu: People in Taiwan. And so, we presented some of the preliminary

results, which included, Oh, sorry, Go ahead.

Abdel Abdellaoui: I was just going to ask; will that be publicly available? The data for other scientists like UK Biobank.

Steve Hsu: Yeah. Unfortunately, it's not going to be as open as UK Biobank. The Taiwanese government is very sensitive about letting the data out. And so really the people working primarily with the data, directly with the data are all physically in Taiwan. And, my group is collaborating with them remotely, but we don't have direct access at the moment to the genomes. but the results are quite promising in that, you know, I, I don't, I, maybe it's for the, I think it's for the first time for a number of conditions, we have predictors which are as strong for predicting disease risks in Taiwanese or ethnically Chinese people as, the west, the corresponding European trained predictors for Europeans.

Abdel Abdellaoui: Oh, nice.

Steve Hsu: So, yeah, so I think, I think it's going to happen. And they, they're, they're scheduled to get into the millions. They're actually recruiting through the healthcare system. They have the electronic medical records and

Abdel Abdellaoui: Yeah. So eastern, eastern countries are doing really well at, catching up with the data, maybe even surpassing at some point. And this reminds me of a Chinese study on major depression. So, they did the GWAS in China, and, and the genetic effects on depression that they saw there correlated genetically only about 0.4 with the European one.

But what I was most interested in was, if you look at, so the genetic overlap with other traits, you saw very interesting differences. For example, BMI and diabetes are positively and significantly positively correlated with the depression genetic effects in, Western countries. So, genes that increase your risk for obesity also increase your risk for major depression.

But in China, the genetic correlation was also significant, but the other way around. And which I can only explain by probably a cultural difference, but I don't know enough about Chinese culture to interpret this finding. So that's a good example of what we were talking about earlier when we were talking about the polygenic scores for major depression and how useful those can be in the clinic. It really just depends so much on the social environment, how your genes become associated with these complex outcomes.

Steve Hsu: Yeah, I think we're going to discover lots of interesting,

phenomena like that. One thing that we reported on in the poster was that we looked at breast cancer risk predictors. We had ones that were trained in European populations, and then we had another one, another set that we're trained in East Asian populations, I think mainly Japan data. And then we tested those on the data that we had through, Taiwan Precision Medicine Initiative. And the interesting thing is that there was not very much fall off, very little fall off, it looked like breast cancer, whether you trained in Europeans or in Japanese, you could predict equally well.

Abdel Abdellaoui: Yeah. So that doesn't seem to be a trait, which, you know, depends a lot on ancestry, LD patterns, or on environment.

Abdel Abdellaoui: Yeah, yeah. There's a lot of variation between trade schizophrenia, which was also done in China. It was almost perfectly correlated with the Western one as well, which had a genetic correlation of 0.98.

Steve Hsu: Yeah, so I think we have examples of both things where, one, where the, within our collaboration we use the word transportable. So, we have some examples of very transportable polygenic risk predictors, and then we have some which are very poorly transportable.

Abdel Abdellaoui: Yeah.

Steve Hsu: So, and the only way to check is to have big data sets in all of the target populations that you want to

Abdel Abdellaoui: yeah. yeah. Did you have, did, did, did they look at educational attainment in Taiwanese?

Steve Hsu: We have not done that yet. We've done it, we've done that particular calculation, I think, and probably you have too, like East Asians that are in UK Biobank, but we haven't done it using the much larger, you know, target data, say Taiwan or something. we just, we're just at the beginning of all of this.

Abdel Abdellaoui: Sounds exciting.

Steve Hsu: It is exciting. I think the future is going to be, and I actually, there's an, you know, this, this is getting into like professional talk, so sorry to the audience, but, but you know, all of us, this US based, data set, is starting to look pretty promising. It had a very slow start, but now it looks like they're going to have a pretty large and diverse, you know, data set. So, I think, I think we made some estimates of how well eventually, if, when all of us is fully complete, how well you could do for various disease conditions and,

Abdel Abdellaoui: A lot of people are complaining about All of Us not being available outside of, for researchers outside of the US.

Steve Hsu: Yeah. And that's the same, you know, the, look, I, I, I agree with you about the Nobel Prize for UK Biobank. It, it's, it's just kind of revolutionized the whole field of genomics and they were so good, so open letting researchers all around the world download instances of the data to do their own analysis. And the U US is not allowing that with All of Us. And Taiwan is not going to allow it with their data. So, it's too bad more people aren't like the British

Abdel Abdellaoui: Yeah.

I, I wanted to come back to something you said earlier by, Sorry, I, I digress by raising the Taiwan thing. But I wanted to get your opinion on longevity.

Steve Hsu: So maybe you're aware of this result. There's this old result which maybe was obtained using something called the Scottish Mental Survey Data. So, I think there was a year or two where they gave them a cognitive test to every kid in Scotland in a certain grade.

And then they did this huge longitudinal study where they followed these kids and, and they're almost all dead now. and so, they know the longevity, the, the, the age of death for each of these kids. And one of the re results, this is through a guy you probably, you know, named Ian Deary. He reported that the cognitive score at an early age for these kids, even after you control for the socioeconomic status of the family of the kid, is a better predictor of their longevity than lots of individual indicators taken late in life.

So even if you knew the BMI of the guy late in life, or you knew the blood pressure, or you, you know, you knew various things. Things late in life about the person you were still better off in making your prediction just using the mental score at age 11 or something of this person. And the only late in life indicator, individual indicator that was more powerful than the cognitive score was smoking status.

Abdel Abdellaoui: Ah.

Steve Hsu: So now obviously the causality here because of what you just described; it could be extremely complicated. Like why does cognitive ability measured at an early age predict longevity? And I was wondering if you had any thoughts about that.

Abdel Abdellaoui: Well, it's, I was a little surprised to hear that it even remains after controlling for socioeconomic status, but also maybe, you know, socioeconomic status is, it is kind of a crude measure probably.

Steve Hsu: I just want to clarify. They weren't controlling for the socioeconomic status of the individual late in life. I think they were controlling for the family that they were born into.

Abdel Abdellaoui: That's okay.

Steve Hsu: Which is different cause I think you're about to say, oh, the high cognitive kids ended up wealthy and living in good neighborhoods with good food and they, they understood how to exercise, what you know, and,

Abdel Abdellaoui: That's part of it. Part of it is being able to get yourself in a position where your life is easier and healthier. Probably part of it is also just being able to make wiser decisions about your behavior. And those two, of course, are also related to each other. So yeah, I, I am not super surprised that those are, are related to each other, that, that that's such a good predictor.

But my guess would be it was a combination of just individual level decision making. So, should I. It's this healthy meal today or go to the fast-food restaurant. But I also think a part of it is just being able to influence where, where you live, like the neighborhood that you live, the place that you work at, the people that you are surrounded with, that's all, strongly influenced by how well your brain works.

I am not too surprised by that. By that connection.

Steve Hsu: Yeah. So, j just to emphasize again that the, the SA SES correction is, I believe not the SES of the individual late in life, but just of the family that they were

Abdel Abdellaoui: Yeah. Yeah, because that, of course, also correlates the, the, the, if you. Born with higher cognitive abilities, chances are that your parents also have them, and they have used theirs to provide a better environment for you. So, I assume that they try to control for that.

Steve Hsu: Yep. But I think that, and, and all of the effects of, okay, if you're above average in your mental capabilities and therefore you're able to get to a point later in life where things are comfortable and, just better for you overall, then yeah. It's not surprising. That increases your longevity.

The other part though, which some people have hypothesized, is that there's also just at the, at the core biological level, within the biological systems in your body, there's some correlation between having good health and having your brain develop well.

So, disentangling those two effects I think are really important because I think it's just currently unknown whether, you know, soundness of mind and soundness of body, how, how correlated are those two things, Right?

So that particular causal pathway, I think if, if one we're able to make the detailed corrections that you could make on the data, you could try to see if there's still this residual which correlates, you know, soundness of brain development with soundness of, you know, your, your health organs and things like this.

Abdel Abdellaoui: Yeah, yeah. Just sort of your general genetic or physical health. Yeah. That would be very interesting to tease out further. Yeah.

Steve Hsu: Yep. Well, that's, I think, a bright future for this subject. I think all these interesting questions are going to be addressed, da, eventually, when we have enough.

Abdel Abdellaoui: Yeah, I hope so. Yeah. Yes, we, I, I hope the data collection keeps expanding at the rate that it is. So now the Eastern, countries are doing really well. I think there's a lot of genetic variation to be won in Africa still, the most genetic variation. And just having a lot of genetic variation measured in many different environmental and social circumstances will definitely, I think, teach us more about how our genetic makeup interacts with the different environments that we build. And then hopefully, that will help us in the future to probably build societies that are more fit for genetic variation in the population.

Steve Hsu: Yeah, you, you could imagine a country which is very progressive or even socialistic, you know, in those countries maybe having a higher EA polygenic score doesn't help you as much in longevity because even poor people get good nutrition and are well cared for. As you know, as opposed to America or something. So, all of that will be extremely interesting. Like FinnGen. You know, the Finnish bio bank. FinnGen finds very different numbers for, you know, the causal pathway, right? If, if you're below average in EA what's the impact on your longevity or something? If they found a very different result, it would be a very strong argument for progressive social policies. You know that Nordic countries have things like that are super interesting.

Abdel Abdellaoui: Yeah, I would be interested in that too. Yeah. Yeah. Because I've only been able to look at Great Britain because today, they are the only ones with a big enough and public data set. But that's like the most unequal country in Northern Europe. I think like the five or 10 poorest regions from Northern Europe are from Great Britain. And the richest are as well. That's in London,

Steve Hsu: Yeah, you should, you should reach out to FinnGen and see if they'll collaborate with you on exactly the same analysis and see what you.

I wanted to ask you about something that recently happened. This is related to data access. And a guy called James Lee, who I think might be a collaborator of both of ours. I think we are both. You and I have co-authored papers with James.

Abdel Abdellaoui: Yeah, yeah. I met him this year. He's a very nice guy.

Steve Hsu: Yeah, he's a professor at the University of Minnesota and, in fact, I think I've, I've interviewed him on this podcast before.

Abdel Abdellaoui: I've heard that episode. Yeah.

Steve Hsu: Ah, good. So, he told me the following story, which is that, and, and he's written in kind of editorial about this, which is that the National Institutes of Health, which is the, you know, US funder of biomedical research, has cut off certain groups of researchers from access to some of their data.

And in particular, I believe it's researchers who study educational attainment. I think not completely limited to that actually. I think Stuart Richey, who you may know who's a British researcher also was denied. I think he had an application which was denied. Um, And he was looking at cognitive ability over time. His interest is in cognitive decline with aging. And I guess NIH in the current kind of sensitive environment, doesn't want to encourage research on some of these, you could call them psychological or behavioral traits. And has said that the reason given I think both to Stuart Richie and to James Lee or James Lee's collaborators who had applied for this access, that the results of their research was potentially quote, stigmatizing. Therefore, they had been denied access to the data. And I think you might actually be impacted by this because you're part of SSGAC, which is the big collaboration that studies team just thought I'd get your reaction to this.

Abdel Abdellaoui: Yeah, yeah. That's, that's not a positive development at all, especially from such a big initiative like NIH, who's, now, our goal as scientists is supposed to be to paint a picture of the world that's as, as accurately as possible, that reflects reality. And we know from the studies that I described, the studies that you described, that things like intelligence and cognitive ability have a huge impact on people's lives, on our society, on people's health. So just ignoring that big part of variation because, I mean, yeah, I'm not even sure why they think that this is something that's development, that is, more pronounced in the US I think, than in Europe where intelligence research has kind of a bad name.

So, I could try and guess why, Sorry. Like in, in intelligence. Is, is, yeah, I guess it's because maybe we put so much value in intelligence in our, in our current society, that focusing on the heritability of intelligence, might the pro, maybe they're afraid that it might, leads to the conclusion that some people are just born less valuable than others. I think that's what they mean, maybe with stigmatizing, which is a, it's a property of our society. We reward people that are intelligent. We give them better rewards. We give them, that's what bio study is basically about. Just ignoring the fact that genetics plays a role in that is, I don't think that's, that's a positive development. We're supposed to try and understand how our society and how human differences work on, on every level. And genetics is just important, a big part of it. And frankly I'm very surprised at this.

So, they think, give any elaboration for NIH, like, do they not believe that genes are important for intelligence, or they just don't want to look at it?

Steve Hsu: It's funny because the institute that controls this data, this is specifically something called dbGaP that I, I know you're familiar with. The Institute is NHGRI. So, it's the Genomics Institute. You know, NIH is organized into individual institutes, and those institutes tend to focus on certain subsets of diseases. NHGRI is specifically focused on genomics in general, I think. And they were the ones who denied these requests for access by both, I think Stuart, Richie, and so James' and your collaborators. so, I don't think they're denying genetic influence on these traits. I think they just feel like, um, I don't know. It's dangerous Or the outcome. Well, they literally say it, these are the results of this research could be stigmatizing.

Abdel Abdellaoui: Yeah. So, but I agree with, with, with, with that part, like we, we should be super careful and it's good that people are critical and keep us on our toes because we have seen it go wrong before, like the whole eugenics’ movement. A big part of that was also related to observations in the early last century. Things like intelligence are very hurtful and people link that to the value of people. And were, were starting widely supported movements that were trying to improve, like the genetic quality of the population because they, but, but. I hope that we can learn from that history and stay vigilant. But that solution is not just stopping this research altogether and, and, and, and trying to ignore genetic influences. I think that's a bad idea.

Steve Hsu: Yeah, I, of course I feel the same way that you do. I think that from the institutional perspective, the way it works is that if you're a bureaucrat in one of these organizations, like NHGRI you know, what's your incentive look like? Like you don't want to be in trouble, right? So, if there are some people complaining, right? Some people read the EA papers and they say, this is outrageous. These people are claiming, you know, genes affect intelligence or something, right? And, and they get some complaints and, you know, maybe, maybe the people complaining or saying, oh, this supports racism or something. And then the bureaucrat can just say, okay, let's be really careful. Let's, let's not, let's stop these. I don't, yeah, I don't like this, EA research. Let's stop it.

But then I think what has to happen then is scientists who don't believe that's the right policy, which is what James is doing. They need to speak out and say, This is not the right way for NIH to handle it. And then gradually those bureaucrats will probably reverse their, I hope, reverse the position on this. I think that's what may happen. Probably you'll be asked to sign some petition. I think SSGAC has a huge number of collaborators across many universities, and probably at some point you'll see a petition.

Abdel Abdellaoui: Yeah. I would sign that.

Steve Hsu: Yeah, me as well. So that, I think that's the one slight cloud on the horizon. But I think it's in a way, kind of minor compared to the fact that, you know, things like the Taiwan Biobank have come online and FinnGen. And as I was saying to you earlier, even within NIH, they're, they're All of Us data set is very diverse. They succeeded in getting a very diverse representation of the participants, and it's finally big enough to really make meaningful meaningful impacts. So, I think the overall picture is quite positive.

Abdel Abdellaoui: Yeah, yeah. I agree. I, I also, have a good feeling about the future of genetics. also, like the past 15 years or so, since GWAS have, have, reached sample sizes to, to detect these reliable associations. Things have moved so fast for so many different fields for the, the, the metic physical, health outcomes have, have, developed so rapidly, like, and there was like so many insights from Covid 19, for example. Within just two or three years they have discovered so much through just those genetic associations. And, you can see how, how fast things can go when you, when you let yourself be guided by genetics. And, and this, I don't think they, that, that, that this revolution can be stopped, anytime soon.

Steve Hsu: No, I don't think it can. I, at, at the risk of telling you stuff you already know, but maybe my audience would be interested in it. I could tell you a little bit about what's happening with breast cancer. So, everybody is familiar with these BRCA variants that have been known for, you know, many decades. are specific rare variants that predispose women to breast cancer. But they only are present in typically one per thousand or few per thousand women. So, it's pretty rare in the general population, but they have a pretty devastating impact. They make you much more likely to have breast cancer. And genetic screening for BRCA is already very common. So adult women would be screened potentially for BRCA and if they are a carrier, then they would get early mammograms or early treatment to detect breast cancer as early as possible.

But now, the polygenic breast cancer predictors that I mentioned to you just a little while ago, those, those now are pretty strong. They've been developed, as I said, now they've been developed and validated both in European and now East Asian populations. There are many, many more women who are high risk for breast cancer due to the polygenic effects, then due to BRCA. So, it's a much larger addressable population.

And already the company Myriad, which you may recall, Myriad had the patent on BRCA. They were allowed by the US to patent a gene. They patented the BRCA gene. Now, that's since been overruled by the Supreme Court, but, but that company has already, they're the number one supplier of breast cancer risk testing originally based on BRCA. But now they have rolled out a polygenic version. They're testing incorporates polygenic effects as well.

Abdel Abdellaoui: Yeah.

Steve Hsu: That's, that's becoming widespread, and it'll have clinical, real clinical impact. And one of the interesting things about this Taiwan project is because it's based in the medical system, they're going to try to roll out clinical applications of these polygenic predictors, you know, sooner rather than later. So, I think that's the next dramatic step that we're going to see.

Abdel Abdellaoui: Yeah, it sounds good. That will save literally millions of lives. Probably

Steve Hsu: Yeah, absolutely. I mean, I mean, the number of, it's an order of magnitude, more women who are high, equal as high risk as BRCA carriers, but for polygenic reasons. And so that set of women are walking around without knowing they're high risk, but now we can identify them pretty easily.

Abdel Abdellaoui: Yeah. Great.

Steve Hsu: Yeah. So, so we're, we're well over an hour. I don't want to take too much of your time. Do you want to make any final comments before we conclude our conversation?

Yeah, I think we've covered a lot. You mentioned rare variants for breast cancer. Rare variants are also still a big unexplored area for a lot of complex traits. There's still, I think there's, like 79% of all variants that they, that vary in a population only have, have a frequency of less than 1% or less than. So there's a lot of individual differences that can still be explained by rare variants. So that's another venue for the future where there still lies a lot of possibilities.

Abdel Abdellaoui: So, yeah. I share your excitement about the future of genetics, and I enjoyed talking to you about it.

Steve Hsu: Great. well. After a little while, I'll probably try to have you back on and you can tell us about some more exciting results.

Abdel Abdellaoui: Sure. Yeah.

Steve Hsu: All right. Thank you very much, Abdel.

Abdel Abdellaoui: Thank you, Steve.