 Well, it's a real pleasure to be here. And I want to thank you for sticking to the last talk after not having much of a break. And I also want to thank my co-presenters in this session. It's a very exciting session. Hopefully the slides will come up in a moment. So if we take a step back and think about the last, say, five to six years of complex trait genetics, one can say that they've been extraordinarily successful in identifying common genetic variants that are associated with some amount of disease risk in populations, largely of Northern European descent. And so hopefully, is that not coming on? OK. That's your guy's laptop. That's not my laptop. So if you pass me my laptop there, never good. OK, thank you. So as I was saying, if we take a step back and think about the last, say, five to six years of complex trait genetics, they've been extraordinarily successful in identifying common genetic variants that are associated with some amount of disease risk, largely in populations of Northern European descent. And this is kind of a quintessential example of the so-called genome-wide association studies. This is a study by my former post-doctoral advisor, Peter Donnelly at Oxford, which was really a watershed moment in complex trait genetics. This was published in 2007, where they took 14,000 cases, so 2,000 cases for each of seven different disorders, bipolar disorder, coronary artery disease, Crohn's disease, hypertension, rheumatoid arthritis, type 1 and type 2 diabetes, as well as a set of 3,000 controls and produced so-called Manhattan plots. So the y-axis here is the strength of p-value of association between a particular genetic marker and disease outcome. And the x-axis is positioned along the genome. They're supposed to remind you of the Manhattan skyline. And each of these towering skyscrapers is a region of association with each of these different phenotypes. And this is really fantastic and wonderful if you're British and born between the months of March and April in 1958, or a British blood donor. Then the welcome trust has invested millions and millions of pounds, which is more than dollars, to figure out why you have a particular set of genetic variants that may or may not be associated with disease outcome. But what if you happen to be born in Cairo or Caracas, Caraccio, Washington, DC, are any of these variants relevant to your population or to why you may have a particular disease? And may you or your population harbor a different set of variants that won't be characterized by this kind of a study? And as Esteban mentioned, one of the things that we are particularly concerned about is the really lopsided investment in populations of European descent. And the consequences that that might have. And I'm happy to report that we're actually now part, we and others are part of large endeavors to do multi- and trans-ethnic studies that try to get at these issues and understand the contribution of both common and rare genetic variants to disease. And that's what I hope to tell you a little bit about today. But really, we need to understand some very fundamental questions about the structure of human genetic variation at this scale in order to be successful in this enterprise. So we need to understand to what degree are human populations that are going to be enrolled in such medical genetic studies substruction and different from one another? What are accurate models, statistical models, for describing this sort of variation within the confines of doing medical genetic studies? How should we or can we use these kinds of genetic data to reconstruct personal genetic history? And why might that be important in this context? And what does it tell us about the design of future medical genetic studies? So one of the projects we've been particularly involved with and I'm very happy to be involved with the Southern Genomes Project, Irvinda has also been involved with this and Esteban as well. Esteban and I in particular worked on some of the collections in the Americas and thinking in particular about the impact that Admixture has on patterns of variation. And I just want to give you one result from this study that I think is really important in our thinking about the next, say, five to 10 years of medical genetic studies. And this is the only bit of population genetics in most of the talk. This is an esoteric summary of genetic variation known as the joint allele frequency spectrum. So if you hold on for a minute, hopefully, you'll see why this is important. So on this axis, we have the frequency of genetic variance in the Yoruba from Ibadan, Nigeria. On this axis, we have the frequency of genetic variance in the Chinese and Japanese sample. And all we've done is cross-tabulate and ask, when we find a genetic variant, what box does it fall into? And the summary of this slide is that the vast majority of genetic variants that is individual genetic changes in the human genome are, one, exceedingly rare. And you actually predict this from population genetic theory. So most genetic variation is actually not common. Common variants are rare. And rare, common, variants are very common. And when we're thinking about variants in this frequency spectrum, that is, those that are between 1% to 10%, and in fact, even below 1%, they tend to be extraordinarily population private. So what does this mean? If we are now delving into these sort of variants, this is where a tremendous amount of the effort is going, to catalog associations between genetic variation and disease, it really, really matters what populations are enrolled. Because we have an a priori expectation that associations found at this scale will not translate across ethnic groups. And Esteban and I had a commentary on this in Nature, which basically asked a very simple question. If you take a pair of populations, say a European, a pair of European populations, or a European and a Chinese population, or a European and an African population within the 1,000 genomes paradigm, and ask, to what degree are they exchangeable in the sense that you'll find matched individuals? Well, for common genetic variants, populations are extraordinarily exchangeable. These are mutations that have been around for tens of thousands of years. You're going to find them all over the world. Pairs of European populations look indistinguishable, and you expect replication across Chinese and European populations, or European and African populations. When we're talking about rare variants, and again, this is the bulk of human genetic variation. Most mutations are extraordinarily rare in population private. You have no such expectation. So now the under-representation of ethnic minority groups in the United States in such studies is an extraordinary problem that needs to be addressed, because if we do not participate, we will not benefit. I will repeat that. If we do not participate, we will not benefit. Okay, and let me give you one concrete example of rare genetic variation and why it may be medically important. And this is really the pioneering work of Helen Hobbs and Jonathan Cohen. This is an extraordinarily important gene called PCSK9, and what Hobbs and Cohen did is resequenced this gene in the Dallas Heart Study, and they found an extraordinary result. This gene, when knocked out, reduces your baseline levels of LDL, or so-called bad cholesterol. So they took 3,000 subjects in this case, and for those individuals that had mutations in this gene, about 3% of them carried knockout mutations of this gene. They had a 28% natural reduction in LDL. Okay, these are not people on medication. They naturally have a lower LDL. Here's the distribution of LDL, and here is the shift in LDL for individuals. This is particularly for black subjects, and this is true in black subjects, as well as white subjects. These variants, again, are population private. The effect is the same. The biology is the same, but the set of variants are gonna be different across the two. What's actually most important about this result, and it's an absolutely beautiful result, is that these mutations not only lower your LDL, but they have a noticeable impact on coronary heart disease. So number one establishes a direct causal link between this gene and the adverse medical phenotype, and it establishes that there are gonna be differences across populations in the actual set of variants that are involved. Now, we're all going to benefit in the sense that now five different drug companies are making monoclonal antibodies to knock down this gene so that we can all benefit by taking a PCSK9 inhibitor, and some will argue that much like statins, you probably, you know, they should just put this in the water, right? Cause it'll reduce our LDL, and here is sort of the biology that leads them in this direction. So there's a very nice example about rare genetic variations, causality, differences in terms of the distribution of the genetic variants, and the need to really think in terms of broad representation in medical genetics. Okay, so the upshot of this is that I really believe, and I've been working towards trying to document examples of where these sort of rare genetic variants may be common somewhere, and this is sort of a map of places where we've been involved in projects and sort of active field work, and you know, I want to sort of take this out, I'm actually not going to tell you about ancestry, you know, I really sort of took the remit of focusing on phenotype, but one of the, you know, very, you know, consistent messages that has come out is that if you look here, for example, these blue pinpoints are all the pop-res samples which gave us such exquisite resolution within Europe, and some of you know we're involved in an ancient DNA project with a guy named Utsi, and we could pinpoint Utsi's ancestry to Sardinia, you know, within tens of kilometers, here's this 5,000-year-old mummy, and we could tell him exactly where he's from, but for most of us in this room, we can't get anywhere near that kind of resolution, so I'll ascribe to something called the Utsi rule, which is that we should be able to do for everybody what we can do for Utsi, okay, and I'll just sort of my aside on that. And the story I want to close with is actually this issue of why these rare genetic variants may have phenotypic impacts, and if we really think hard about phenotype and about sampling, how we can actually learn some pretty cool biology. And to me it's actually a very fun project we were involved with, and some of you who were in the exhibit is actually in the exhibit two floors up, and it's the case of blonde hair in Melanesia, okay, so this is a project that I was sort of unwittingly involved with in that, this postdoc in my lab, Sean Miles, came to me with this photo, and he said, Carlos, I want to study the genetic bases of blonde hair in the Solomon Islands, and I will admit, when I first talked to Sean, I had no idea where the Solomon Islands are, for those of you who don't know either, the Solomon Islands are a sovereign nation off the coast of Papua New Guinea, and when he showed me this picture, I said, well, Sean, look at this kid's US servicemen's jacket, I can probably tell you the genetic bases of blonde hair, his dad had blonde hair, it's a sort of Captain Cook allele, and he said, no, no, no, no, I think it's a different, it could be potentially a different gene, I said, oh God, we've really studied blonde hair a lot, so it's okay, but if you get a grant and can get the samples, then we'll think about it. So Sean went out and he got some funding, and went out to the Solomon Islands, and here's Sean with spectrophotometric, taking spectrophotometric measures of kids in the Solomon Islands, this is a project we did while I was back at Cornell, we collected about 1,000 subjects and measured their hair, and there's actually a lot of interest in the Solomon's, if you look at the Lonely Planet Guide, that's actually, there's a blonde Melanesian kid who's featured on the cover, so, and they actually had no idea why the genetic base of blonde hair, they had all kinds of different theories about it, and so we were interested in thinking about the problem. So then he came back to the lab at Stanford and said, I'd like to do a genome-wide association study, and I said, well, at this point, Sean, this is really, really expensive, we don't have thousands of arrays that we can throw at this, and he said, no, I think if we did it with 100, it would work, and I sort of looked at him and said, okay, clearly you've never taken either a course in statistics or looked at a general issue of nature genetics, these are, the one that I told you about before had 2,000 cases and controls, and Sean here then had a very good insight, which is let's focus on the extremes of the distribution, which is exactly what Helen Hobbs and Jonathan Cohen had done, they had initially sequenced individuals from the tails of the distribution. So he said, fine, let's genotype these kids with blonde hair, these kids with dark hair, and then I said, you know, when this doesn't work, Sean, you really need to go work on your other project, and you know, so you can sort of get out of the lab. Okay, so these are the results, so this is what we like to call our Dubai plot, not our Manhattan plot, because there's a single towering peak of association. It's on chromosome nine in a region called P23, and in fact, we got extraordinarily lucky. Within this region, there's actually a single gene that is associated, and it's not some predicted hypothetical gene, it's an actual real gene with honest to goodness, you know, research that people have done on this gene. And in fact, the association has an odds ratio of 30, which I challenge anyone out there, I think it's still the largest odds ratio with a GWAS. So, when you delve into it further, in fact, it's a gene called Terp1, tyrosinase-related protein one, which is expressed in the melanocytes, so it's in the cell types that produce melanin. It's involved in melanin synthesis and in the maintenance of the melanosomal structure. It affects cell death and proliferation, and fascinatingly, there are cases of people with knockouts of this gene that have a form of albinism called Oka3, or Rufus albinism, or albinos of dark skin. The first family that was described to this was actually a South African family who were essentially dark-skinned parents with albino children. At this point, as a good PI, I became very interested in this project, and green-lighted that we do some additional sequencing, and we sequenced the gene, figuring that we're gonna have a ton of markers that we're gonna then have to delve into. It's probably gonna be some non-coding element that's involved, these GWAS events are never clear, but again, sort of we were extraordinarily lucky. We had one and only one mutation, and it turned out not to be just any mutation, it was an amino acid change. It's a change from the amino acid arginine to the amino acid cysteine, an extraordinarily conserved position. So here's the protein sequence, and these are all Rs, which stand for arginine, and I'll just read these out to you for dramatic effect. This is human, resis, dog, elephant, possum, out to chicken and zebrafish. So all of them have an arginine at that position, and our kids with blonde hair are walking around with a cysteine. Furthermore, we had 1,000 kids, so we could actually test directly for association, and here is the distribution of hair pigmentation. Here are kids that are homozygous, homozygous arginine, here are the kids who are heterozygous, these two distributions are the same, and then our kids who are homozygous for the cysteine mutations are two standard deviations out, okay? So here's a phenotype, blonde hair. There are two populations in the world that have blonde hair, Melanesians and Europeans, so we can now ask, is it the same set of genetic variants, and the answer is no. This is what one might call chemically induced blonde hair variation. But if you're European and have a blonde hair, then you would tend to have a mutation in Ocatoe, or perhaps Ketligen. You do not have this Turp 1B mutation unless you have Melanesian ancestry, okay? So in my mind, I think it's a very, perhaps representative example of what we would learn if we broaden ethnic representation in such studies and focus on phenotypes. I mean, it's easy to phenotypically identify blonde Melanesians so they make a ton of sense, but if you imagine being able to better phenotype, you could then dissect the genetic basis of these complex traits with perhaps smaller sample sizes. So the conclusion here is that even traits like hair and skin pigmentation that are highly heritable, we have yet to identify all the genes involved. We do not know whether, for example, the dark skin pigmentation in these individuals from Melanesia has the same genetic basis as dark skin pigmentation in other parts of the world. Other complex traits may similarly be amenable. We need to undertake, in my mind, phenotyping across diverse human populations. To me, it breaks down any simple notions of race. The alleles are geographically distributed and contribute to phenotypic differences. I think in this regard, I'm in a violent agreement with Eravinda that this isn't about race and classification. This is about the distribution of alleles that have impacts on biomedical traits. And I think it's really, really important that we characterize this. I'll leave you with two final thoughts, which is that this gene, number one, has a mouse knockout. So we actually have good biological proof that this is the gene that's gonna be involved. And secondly, it is the gene that is most overexpressed and metastatic melanoma, right? So there could be a direct potential medical link there. So with that, I'd like to thank you all, thank the people of the Solomon Islands. It's a privilege and pleasure to be able to work with them and happy to take any questions you may have.