Genes and Intelligence: A Mensa Foundation Prize Winner Discusses Groundbreaking Findings

For her work identifying, for the first time, hundreds of human genes highly correlated to variations in intelligence, statistical geneticist Dr. Danielle Posthuma of the Netherlands was awarded the prestigious Mensa Foundation Prize in 2021.

The Mensa Foundation Prize is awarded biennially for the best scientific discovery in the field of intelligence or creativity.

Dr. Posthuma is head of the Department of Complex Trait Genetics at the Vrije Universiteit Amsterdam and Amsterdam University Medical Centre, where she leads researchers in statistics, stem cell biology, and bioinformatics and conducted two large-scale genetic discovery studies into intelligence. She recently discussed this research with the Mensa Bulletin.

* * *

Chip: So, I’d like to talk about your work in general, but obviously, based on the award, want to focus on the intelligence angle. You’re being recognized for directly identifying — for the first time — hundreds of genes in the human genome highly correlated to variations in intelligence. I want to be sure that we put an accurate framework around these discoveries. What have you discovered in terms of genetics and traits for intelligence?

Danielle: We have actually conducted two genome-wide association studies. I don’t know if you are familiar with the term “GWAS” [genome-wide association study], but it means we are looking at all the genotypes that we have in our genome and then testing whether certain genetic variants more often occur in people who score higher on intelligence tests.

Our first GWAS was also the first time that such a study identified statistically significant associations of genetic variants with intelligence. Previous studies had relatively low sample size and were underpowered and therefore did not report any significant findings. When we decided to do this study, we collected all of the data that were out there and conducted a meta-analysis to obtain a sample that was large enough to allow the detection of locations in the genome that were associated with differences between people in intelligence.

In our first GWAS study in 2017 [Sniekers et al.], we detected 18 different genetic locations that were associated with intelligence. It was seen as kind of a breakthrough because for decades we knew that intelligence was heritable, but it was very difficult to find the actual genes that were responsible for this heritability. In the second study, we increased the sample size tremendously, including more than 200,000 individuals.

After that was published … that was the first of two, correct? After the first study was published, you told Nature magazine that the current genetic results explain up to 3 percent of the total variance in intelligence?

Yes. Now we’re up to about 5 percent, I think, with the latest GWAS, but there are other GWASs out there that didn’t directly measure intelligence, but they measured educational attainment, which is strongly genetically correlated to intelligence. If you would look at that, you could explain more of the variance, like up to 11 percent.

That’s still not a lot, given that we know the heritability of intelligence is about 80 percent. So, from that 80 percent heritability, we now know for about 10 percent where in the genome those locations are that are responsible, but there’s still a lot that needs to be discovered.

What do you mean when you say intelligence is about 80 percent heritable?

That means the reasons that random pairs of individuals, drawn from the population, differ in their intelligence is 80 percent due to differences in their genetic makeup.

We can estimate the relative influence of genes and environment by comparing resemblance in monozygotic and dizygotic twins. Since monozygotic twins are genetically identical and dizygotic twins share on average 50 percent of the genes, their resemblance on intelligence tells you something about how much intelligence is influenced by genetics. For example, if variation in a trait is completely due to variation at the genetic level, then monozygotic twin pairs would be twice as similar on that trait compared to dizygotic twin pairs. Whereas if variation in a trait is completely due to environmental factors, monozygotic and dizygotic twins are expected to show similar resemblance.

For intelligence, the correlation in adult monozygotic twin pairs is about 0.8 and the correlation in dizygotic twins is around 0.4. So it’s half the correlation of monozygotic twins. But for monozygotic twins, the correlation is not 1, but it’s 0.8. So the heritability of intelligence is not thought to be 100 percent, but it’s 80 percent. And that means that if you look at it from a different viewpoint, for example, at the point of view of individual differences in a population, so you could ask why it’s not everybody of the same intelligence, but why do you and the person sitting next to you differ in intelligence score? Then across all of the different pairings that you can make in a population, 80 percent of the reasons why two people differ from each other in intelligence is because they are different at a genetic level. And for 20 percent it is because there are differences in some kind of environmental factor.

Wow. That is amazing that you’ve been able to quantify that.

Yes, and that’s something that we have known for more than a century. When the first twin studies were conducted, intelligence was one of the first measures that people were actually looking at. So, we’ve known for a very long time that it’s a highly heritable trait, but it has been unknown for so long what the genes are that are important for intelligence. So that’s why when we first published our genome-wide association study, people were so excited. The concept was that we could finally dive a little bit into the biology of intelligence.

What do you think the realistic limitations of our understanding of this are? Do you think ultimately, with enough study, we’ll be able to understand basically all of the genetics that are tied to intelligence?

Well, that’s a good question. Intelligence is one of those traits that is highly complex, which means that it’s influenced by thousands of different genes. It’s really unclear how these genes collaborate mechanistically with each other to make a person more intelligent than another person. It might also be slightly different in every individual. So, it’s not the case that people who have the same IQ score always have the same genes that are important for intelligence. It might be a completely unique combination of genes that are important for intelligence that you have and that I have, even though we might have the same intelligence score.

So, as with many other complex traits, it’s still something that we have to sort out. So why is it easier for people who have this particular genetic composition to have a higher score on an intelligence test than for other people? That requires a lot of biological knowledge. Connecting the dots between the different genes, sorting out what these genes are doing in a human cell and how they connect to each other, or how they form protein complexes. That’s actually work that’s still ongoing, not only in our lab but also in other labs around the world.

When you think about this work — I understand that you’re doing genetic work across many fields, not just studying intelligence — but in the field, particularly when looking at intelligence, what is it that you personally really want to find out? What’s driving you right now?

In the context of intelligence, I’ve always been intrigued to know why it’s so easy for some people to learn something, to study, and why it’s much more difficult for other people. For example, some people have a very good memory function, it just comes naturally with them, and other people are always struggling. I’ve always been interested in understanding why this is the case — and whether that may have consequences for the way we optimally give instructions to people. If that’s the case and we could actually change these instructions for certain students at school for one part of the class, you have to learn it in a more visual way for example, or if the other part of the class is better off when they read or when they listen to something. These differences in the way our memory works or the speed at which we can memorize information — that’s always been intriguing for me.

I would really like to understand what the underlying mechanism is — what makes people do something faster and more efficiently compared to other people? And then the second thing is that for many diseases I’m interested in, such as schizophrenia or Alzheimer’s disease, cognitive decline or cognitive dysfunction is also an important factor. So if you understand the reasons why people do not have a disease, but why they still differ in cognitive function, maybe that also helps us to understand the things that might go wrong when you do have a disease where cognitive dysfunction is involved. So yes, that’s what’s mainly driving me.

That’s a good segue into some of your other genetic studies. Can you tell me what, in broad terms, are the other major findings from your work studying these massive volumes of genetic data — Alzheimer’s, insomnia, neuroticism?

The focus of my work is actually that I’m interested in method development because I really like to think about novel ways to look at the same data and also to address new questions that come up.

Looking back at the field of genetics, when I started in this field 25 years ago, all we could do was say, what is the inheritability of the trait? What is the relative influence of genes and environment? But we weren’t able to actually look at the genes. Then came the possibility to do these genome-wide association studies, where we could actually look at the genetic variants. Then the whole field was focusing on pinpointing the genes for intelligence or for schizophrenia or for any trait. And finally now those studies are so highly powered that we are actually able to find those genes, but then the next challenge comes up and that’s: How do we interpret our results?

We have found so many genes for intelligence. What does it actually mean? We can say we found 50 genes for intelligence, but no one really knows what that means, if that’s your single statement. It doesn’t really give you any biological insights. That’s the next challenge, and for each of those challenges, I’ve tried to develop novel methodologies or tools. One of the tools we developed, for example, was to look at the combination of genes. What is their shared function in a human organism? That relies on the assumption that if thousands of genes are important for traits, then, I strongly believe, at some point there must be some convergence at a functional level. At some point there must be a similar function of all of those thousand genes. Otherwise, it wouldn’t make sense that they all contribute to schizophrenia or to Alzheimer’s, for example.

One of the tools we developed helps you to find the point of convergence. That’s one part of my research — to develop novel methods and apply them to psychiatric and neurodegenerative and cognitive traits. Then the other part is to connect genetics and neuroscience. Because once we’ve done our genetic work, then if you really want to understand what’s going on at the biological level, you need to do functional experiments in the wet lab environment.

You have to go to mouse models, for example, or other animal models, or you can use stem cell biology to test some of the hypotheses that come out of our work. I collaborate very closely with neuroscientists, and they set up experiments to test the hypotheses that are based on the results of genetic studies.

The step after that is, once they have set up an experiment, they can show, for example, how perturbations in genes associated with a certain trait lead to cellular and behavioral consequences. The step after that is to go to the pharmaceutical industry, for example, and to collaborate to develop novel therapies or drug treatments for certain issues. I think the ultimate goal is understanding — understanding the biology of multiple diseases.

Can you talk a little more about your methodologies? You developed some novel methodologies to process a vast amount of genetic data.

Two of the tools that many people are using in the field are MAGMA and FUMA. MAGMA is the tool that allows you to look at the combined or the shared functional effect of multiple genes in analytical pathway analysis. You can test whether hundreds of genes that are known to have a similar function are involved in a disease or whether certain types of cells show increased expression of the genes that are important for your disease. The other tool, FUMA, helps you to interpret your results of a genome-wide association study. It combines your results with biological information to facilitate and pinpoint the actual causal variant and gene.

FUMA allows you to use external resources that have additional biological information on those genetic variances to, for example, find those variants that actually change the protein structure or variances that change the level of expression per gene. If you combine that information with your p values, your statistical significance, then that helps you to filter out genetic variants that are not known to have any function. They are probably not very likely to be causally associated.

That used to be work that would take six months because you had to sort out all of the external databases, format everything in the right way, then combine it. We created this tool first for use in-house, but then we just put it out in public so everybody else could also benefit from it, and now it takes 30 minutes or so. I think all of those tools help people to understand our genetic findings and gain more biological insights to what genetic discovery studies show.

Interesting. So I’d like to talk about some correlative traits or traits correlated with intelligence. I’m going to say something that is probably going to be a gross oversimplification, but perhaps you could kind of walk me through correcting it. Someone who’s got those genetic traits for intelligence: They are also genetically more likely to be overweight and have mental diseases. Is that correct?

Are you referring to a table in one of the papers or …?

Yes, it was related to a story in Science from 2018: Hundreds of new genes may underlie intelligence but also autism and depression. It also gets to talking about traits for body mass and things like that. Could you kind of talk me through what sort of these correlative traits are and how that works?

I’m looking at that first GWAS paper, where we have figure 3D.

Figure 3D shows genetic correlations. First of all, genetic correlation, it doesn’t mean any causal relationship. It’s just a correlated thing. It doesn’t imply that a higher intelligence is causing lower autism or the other way around.

But, if there’s a genetic correlation, then that means that the genes that are important for the first trait are overlapping to a certain extent with the genes that are also important for the second trait. So we don’t know if those overlapping genes are influencing both of these traits or whether a gene influences trait one and trait one causes trait two. This would also induce correlation. We don’t know that.

That’s an additional analysis that needs to be conducted to sort it out. That’s very difficult to do if you don’t have longitudinal data. There’s no proven causal relationship at this point. That’s one. Then, if you look at this figure 3D, the red bars are significant after correction for all of those different tests that we did, and the other ones are basically not statistically different from zero, or no correlation. Then for example, the first one, that’s a negative genetic correlation. So that means that the genetic variants that are associated with a higher intelligence score are also associated with a lower risk for Alzheimer’s.

Oh, with a lower risk for Alzheimer’s.

Yes, because it’s a negative correlation. And why is this the case? I don’t know. We would have to conduct experiments to really sort out what is the actual course or route here. But, this actually fits with the hypothesis that is prevalent in Alzheimer’s research of cognitive reserve. That means that people with a very high intelligence score have higher cognitive reserves and therefore may be less likely to suffer from Alzheimer’s disease or to have visible cognitive decline or might start just at the later age, because they start from a higher cognitive functioning.

But that’s a hypothesis. That still needs to be tested. Then you ask for body weight, for example. Here you see body mass index in adulthood. It’s sufficiently significant, and the correlation is minus .11. So again, this means that the genetic variants that are associated with a higher score in intelligences are associated with a slightly lower body mass index. It’s statistically significant, but it’s a very weak correlation. This is point 11. You would have to square this to know how much of the variance each trait explains in the other traits. It’s like 1 percent of the variance in body mass index is shared with the variance in intelligence. So it is a very weak correlation, and there are lots of other factors that are important for these.

Again, here, we don’t know the direction of causation. We don’t know what is causing what. For Alzheimer’s you also have to square it out. That’s about 10 percent of shared variance between the two traits.

Our time is nearing an end. Is there anything we haven’t talked about in terms of your research? I know we focused on the intelligence angle, but are there any other things you want to be sure we cover?

I think intelligence research is always a bit tricky. There’s always a lot of sensitivity, and I just want to stress what you already asked, that the reason why we conducted this kind of research is to gain more insights into the underlying biology, and to just understand what’s going on or if can we understand things at a mechanistic level.

Maybe it’s too complex for us to understand, eventually. Maybe some genetics can help us find biological pathways that can help explain why some people have a better memory than other people, or why their brain just works in a different way.

And I think that might be very useful for many people.