So the title of my presentation is 'Who We Are and How We Got Here: Ancient DNA as a Window into Human Biology and History.' And I wanted to begin by thanking Jim for two gifts. The first is support that we received in 2014 and '15 for a project to sequence the genomes of people from more than 150 populations around the world. This continues to be a reference data set that is used around the world — the Simons Genome Diversity Project. And the second is Nick Patterson. So Nick Patterson is my scientific partner for the last 25 years. I met him in 2001, and he's a mathematician. I don't know where my work ends and his begins. And he worked for Jim from 1993 to 2001 before moving to biology, and worked also in cryptography and at IDA before that.
So the talk plan here is: I'm going to talk about who I am and an introduction to ancient DNA, the field of ancient DNA. Then I'm going to talk about four surprises from the deployment of this kind of amazing and startling ability to look at the past through ancient DNA. Then I'm going to talk about the beginnings of what we can do now, which is to learn about biology in ways that are maybe realizing the promise to learn as much about biology as about history from this methodology.
So, I'm going to begin by talking about the first half of my own research, which is mathematical and statistical methods development. And Nick Patterson and our group have really built what's now become the standard toolkit in a lot of ways for looking at human population history. This is a paper we published in 2006 about the application of principal component analysis to studying genetic structure. This was with my first postdoc, who was also a Renaissance employee who had been hired by Nick — Alkes Price — who had been at Renaissance for five years. Then we followed this by developing a whole suite of methods to look at human history with genetic data.
The second half of the work that I've been doing has been generating and analyzing large amounts of genetic data. And we built an ancient DNA research group here in the USA beginning in 2012–13. And this methodology starts with a bone or a tooth, and in a clean room like the one we have, we try to get beneath the surface of the bone and not to contaminate it with the people who are handling it. We take a small amount of the sample. We take some powder from it, a very small amount. We then release the DNA in an optimized solution that releases the proteins and the contaminants that prevent the DNA sequencing from going forward. And then we sequence it.
And often we use a number of pre-treatment methodologies that make this more efficient, cheaper in one way or the other. And as a result of these innovations, there's been a huge explosion of data that has occurred, from the first five genome sequences that came out in 2010 to now more than 20,000 reported individuals with this type of data. And it makes it possible to ask and answer questions that were not possible to answer before.
Another thing that we've done in our group is we've ended up compiling a lot of this data into a homogenized format and serving it out over what we call the Allen Ancient DNA Resource, because it was supported by grants from the John Templeton Foundation and the Paul G. Allen Frontiers Group. So I wanted to begin by talking about four surprises that have come from the ancient DNA work, and I'm going to emphasize ones that I've been involved in — but many I haven't been involved in, but I can speak with a little bit more directness or authority or knowledge about the ones I'm going to tell you about.
So, the first is regarding some of the first genomes that were published in 2010. In 2007, I got really lucky, and I had published a paper about humans and chimpanzees and gorillas and how they're related to each other deep in time. And the work was trying to understand mixture that seems to have occurred in the deep past. And Svante Pääbo, working in Germany, was clearly successfully sequencing DNA from archaic Neanderthals, who are distant, distant cousins who separated hundreds of thousands of years ago from our lineage, who lived in Europe and Western Asia. And he invited us to analyze the data, and in particular to look at the questions about whether they had mixed with us or our ancestors in some ways, as Neanderthals — as modern humans exploded outside of Africa and the Middle East after 40 to 50 to 70,000 years ago. And we know that they overlapped.
Here's a map from a popular science article about where they overlapped. And we know Neanderthals and modern humans for a few thousand years in Europe were living side by side and learning from each other, even. And a question was: did they interbreed? And does that interbreeding — has it left an impact in people living today, especially in Europe, which is where most of this contact might have occurred? And so what we did is we looked at the data. I came from a tradition in my work where we just were — we didn't think that this had happened. We had looked at lots of modern genomes, and there was no trace of archaic DNA that looked at all convincing, and we thought it had been a complete population replacement.
So what we did is we took a few different techniques, but this is a cartoon of a chromosome. And the cartoon of the chromosome is showing a variant that differs between French people and Nigerian people. For example, some people have a DNA letter thymine, some people have a DNA letter guanine. And we just ask how often the Neanderthal matches one more than the other. And what we saw was a very severe, strong excess of matching to Europeans than to Africans. But this was not just Europeans. It was all non-Africans. So also Native Americans or East Asians just as much.
And by asking how much is the skew compared to what you see if you replace it with a second Neanderthal, you see it's 2% of the way toward a Neanderthal, and this directly translates to an estimate of how much Neanderthal DNA there is in the French person or the East Asian person. And the amount is just as high in East Asians and in Native Americans and people from Australia as in Europeans, which is also a surprise because the Neanderthals lived in Europe, so you might think that there's more, but there's not.
The next thing that we tried to do is we tried to build — and we did build — maps of where in the DNA of each person we were analyzing the bits of Neanderthal ancestry occur. And this map is supposed to be showing that — let's see, how does this work? The map is showing this, and you can see bits of the DNA — a person has a bit of Neanderthal ancestry over — it's only overall it's only 2% — but it occurs in spiky ways at some places in the genome. And so you can pile this up in large numbers of people, and above the axis is the proportion in a large number of European people, and below the axis is the proportion in a large number of East Asians.
And you could see where it's much more than the 2% average — there must be natural selection for genetic variants that have been perhaps advantageous in the places where Neanderthals were already adapted and modern humans were moving into. And you also see these deserts of Neanderthal DNA where there's no Neanderthal DNA. And these deserts, we were able to show, are places where there must have been something very important in modern humans that was preserved after Neanderthal DNA was pushed into the population. Neanderthal DNA was removed because that was particularly important for the adaptations of this group.
We also see on chromosome X, one of the two sex chromosomes — two women have two copies, men have one copy — that there's a very extreme deficiency of Neanderthal ancestry. And we know that chromosome X harbors many genetic variants that are important in male fertility. And if you have a mixture of groups, species that are quite different from each other, genetic variants from the minority population are removed usually. And that's also true in genes involved in meiosis, especially male meiosis. And we see both signatures. And so it's likely that Neanderthal–human hybrids had a low rate of fertility, at least on the male side. They were far enough apart for that to happen.
The second thing I wanted to talk to you about was the Denisovans. So this was an incredible piece of luck that, again, we had — Svante Pääbo's group in 2010, while we were finishing up the Neanderthal project, got a sample of a pinky bone from a little girl who had been found in a cave in southern Siberia. And when we sequenced it, it turned out to be neither a modern human nor a Neanderthal, but something else. So if you look at the whole genome and line it up, it's hundreds and hundreds of thousands of years diverged from Neanderthals at the top there, and also slightly more diverged, but many hundreds of thousands of years, from all modern humans.
If you look at the mitochondrial DNA sequence — what you get from your mother and she gets from her mother and so on — it's a million years diverged. And Neanderthals and modern humans are much more closely related, maybe 500,000 years. And then if you look at the Y chromosome, that just came out a few years ago, it's the same thing. It's maybe 800,000 years for the Denisovans and 400,000 years for the mitochondrial DNA. So it's a really special, different thing. And let's see, how does this go?
So the next thing we did is we tried to test for mixture between Denisovans and different modern humans. And here we got this really shocking surprise. This is something that Nick Patterson found. And we looked to see whether Denisovans were more closely related to New Guineans — people who live close to Australia, on a large island — or Chinese people. And we saw a huge excess of matching to New Guineans. And by the same trick, it's evidence of 3 to 5% of the DNA of people who live in New Guinea and also, we now know, Australia and places that mixed with New Guineans is coming from distant cousins of the Denisovans from Siberia, what we now call southern Denisovans.
Here's the proportion of this Denisovan ancestry throughout the Old World. You see there's very little of it in Europe and in Africa, but if this full circle represents 3 to 5%, there's maybe a twentieth of that throughout East Asia. So it's another group of humans, not Neanderthals, that in some places left an even bigger impact than did the Neanderthals.
We now know, as of two weeks ago, what these people looked like. All we had was a little finger bone and some teeth, all from Denisova Cave. But a former student of mine, now working in Beijing, Qiaomei Fu — she got DNA from the one tooth remaining from this skull that was found in a well in the 1920s, and scraped some dental calculus off of it and got a little DNA out of this dental calculus. And it had characteristic sequence changes that match the Denisovans, and also was able to do some protein sequencing which also matches the Denisovan. And so we now know what their morphology looked like and can begin to tie them to morphology and archaeology.
Surprise number three relates to something more recent in time but also very impactful, which is how ancient DNA reveals something very startling about the origin of the people who spread the languages we're speaking now — English — which are related closely to languages from Iran into North India and Armenian and many other languages from Europe. So I'm going to begin by talking about a little personal history.
So I was working on this Neanderthal genome project and Denisovans for Svante Pääbo as an analyst analyzing the data, and there was a faculty search in a department at my institution that I was invited to be a guest on, which was supposed to find a human evolutionary geneticist. And for me it was very clear that we would want to use these techniques to look at human history, not just archaic humans, but also try to understand how the people around the world got to be the way they are today. And that's a question of the last 15,000 years in many ways.
And when this happened, the faculty search committee — we got these amazing candidates, two amazing candidates in ancient DNA, one of whom now heads a research institute in Europe that's one of the most exciting places to do this work. And they determined that these people weren't even allowed to be considered because it wasn't relevant to human evolution. And I don't get angry very often, but I became so angry and I just like stomped out of the room and I said something very bad. And then it really became clear to me that no ancient DNA expert would be recruited.
So I actually just dropped what I was doing. And I didn't have any money at the time for this purpose. And so what I did is I used techniques that were used in medical genetics to make this process much more efficient. And I came out of a genome institute background. So we sort of industrialized the process and made it about 30 times cheaper. And so as a result we've been able to generate a lot of data.
So the Indo-European origin problem. It was observed more than 200 years ago that languages like Sanskrit and most northern Indian languages like Hindi, or Iranian languages, are closely related — probably within four or five or six thousand years — to languages of most of Europe, except for Hungarian and Finnish and Basque. And this must be because there was some common ancestral population that lived somewhere. And the hunt was on to try to figure out where that was until 2015.
The primary view was that this was due to the explosive spread of agriculture out of the Middle East, which agriculture is invented for the first time anywhere in the world about 12 to 11,000 years ago. And it spreads into Europe about 8,500 years ago, at the same time as it spreads eastward into Iran and eventually ultimately into India. And we know how it spread. It's a complete economic revolution. And the thought was that this would have spread these languages. So then the ancient DNA revolution happened, and we found that that really was unlikely to be the case.
So here is a principal component analysis of genetic variation from about a thousand people from the places marked in the map on the right, from Western Eurasia, Europe and Western Asia. And each dot is visualized — there's a matrix showing how closely related each individual is to every other individual based on its genetic variants. And they're colored by where they're from. And you see an amazing pattern when you do this to modern people. All the Europeans are on the left. All the Near Easterners are on the right. There's almost nobody in the middle. And that's because of some distinctive ancestry that separates these two groups of modern people.
We now know a lot more about this and how it's changed over time. But what I'm going to do is gray out those dots and show how the ancient individuals that we now have sequenced from, and that we began to get sequence from in 2014 and '15, relate. So the ancient individuals, the hunter-gatherers of Europe who lived more than eight or nine thousand years ago, they fall beyond Europe in the direction of European differentiation from the Near East. And the only way that that can be the case — and we've shown this in multiple ways — is if Europeans today are a mixture of those hunter-gatherers and some kind of Near Easterners most closely related to Anatolians.
Anatolian farmers beginning 8,500 years ago appear in Europe in a very different place and spread into the Balkans and into parts of central Europe, and they fall at the bottom of the current European gradient. Meanwhile — so that's the result of this explosion from Anatolia into Europe. So meanwhile, in the steppes north of the Black and Caspian Seas, what's now Ukraine and western Russia, these people appear. These are the Yamnaya steppe pastoralists. These are the first people who take wheeled vehicles out to the open steppe, and they spread in a very dramatic way all the way from Hungary to Mongolia after this.
These people appear at the top of the European gradient, but you don't yet see people who look like Europeans by 5,000 years ago. And then by 4,500 years ago, boom, modern Europe appears. So something very dramatic happens in Europe, and it's due to the mixing of these Yamnaya with these Anatolian farmers. So that's this second migration. And they mix after 5,000 years ago. And so these are almost certainly the Indo-Europeans, and they're spreading Indo-Europeans. The Yamnaya are almost certainly the people who spread late Proto-Indo-European languages. I'll talk about that a little bit more.
First I'll talk a little bit about what happens as these people move west. So this is the proportion of steppe ancestry, Eastern European ancestry, in Britain beginning 6,000 years ago. So farming first gets to Britain 6,000 years ago. Before this there are hunter-gatherers, and it's a 99% population replacement. This is very dramatic. And I'm not showing you that until 4,500 years ago nothing changes in terms of the ancestry, more or less. The last big stones at Stonehenge go up, and then within a hundred years — a hundred years — there's a huge population replacement in Britain. These people sort of spread into Britain within a hundred years. There's almost nothing left of that population. We now know they come from the Netherlands, where they formed rapidly within about 100 years before this.
Here's the same time course in Iberia, in Spain and Portugal. Same time period. And you see the steppe people come in at the same time. The colors of the dots are the males, and they're colored by whether the Y chromosome that they inherit from their fathers is typical of Eastern Europe or not. And what you see in Iberia is that the mixture is only 40%, not 90%. But you see that there's almost no males left from the local population. So what happened is that the local males almost completely failed to reproduce, even though it's only a 40% mixture. So these are sort of dramatic social turnovers that you're seeing evident in the data.
So in a whole series of studies, we and other people have traced the dilution of this Yamnaya ancestry as it moves west into Europe, hitting Britain as I showed you, or hitting Iberia and other places, and then with this backward movement over here, twists around and goes back and punches through the mountains and goes into India after 3,500 years ago or after 4,300–5,000 years ago. That backward movement corresponds to the genetic high similarity of people in the Baltic, like Lithuanians, and South Asians and Iranians, who also speak related subgroups of Indo-European — Balto-Slavic and Indo-Iranian languages. And so we can trace the movement of these languages through this process.
It's not always the case that culture matches genes. So we just recently published, as an example of this, DNA from ancient Carthaginians and people from the Punic world. And what you can see is that the Phoenicians, who are supposed to be the parent culture — and they are the parent culture of the Carthaginians — are over in that circled area of the principal component analysis near previous Canaanite peoples. But the Carthaginians and the Punic people have no overlap with them and instead are genetically largely Greek or North African, and mostly Greek.
And so even though they spread culturally, it seems to have been adopted by the other people they were interacting with — the Greek colonial expansion — and somehow these are Greeks who became Phoenician through adopting the culture, just like many people in New York have adopted aspects of the culture even though they're not ancestrally. And that's something that began to happen for the first time in the European world with the Phoenicians. It's kind of, in some ways, the first cosmopolitan culture.
And the first surprise I wanted to tell you about was the ability to look at kinship in the deep past from ancient DNA. So with these huge numbers of people, you can begin to excavate whole sites. And this is an excavation of a tomb that was excavated in the 1970s from Gloucestershire in southern England, from a tomb that was about 6,000 years ago. This structure is about 50 m long and 10 m wide. And in the middle are these two chambers filled with bones. We sequenced whole genomes from 35 people. And 27 of them were linked in this amazing pedigree that we can work out perfectly.
There is — squares are males and circles are females. And you can see a number of patterns here. One of them is that males and females are both reproducing with multiple partners. And so you see, for example, that square male at the top reproduced with all four of the females at the top, but the two left females over there both reproduced with other males who then produced children who are buried in the pedigree.
Another is that there's extreme patrilineality in this community. Every person who you see can trace their line through male ancestors to the beginning. So this person traces to this, then traces to that. This person traces this to that to that to that. And so it's an extreme patrilineal community, but it's actually not, because there's cases like this one that are cases of adoptive kinship.
So this person you think could trace his ancestry through his mother, but actually what happened is his mother had another child, this one, with someone who actually — with the man in the lineage. So the rule in this family seems to be: if you're the son of a woman who has another male son with someone in the pedigree, you're valid and you get to be included in the pedigree, and this son can be included. The same is true of this one, where this person is only included because his mom had another kid, and so on with that one over there.
So this is observed also in some communities in, for example, the Nuer in Sudan, where this adoptive kinship occurs in the context of expansion of communities as a way to grow families rapidly. Finally, you also see matrilineal communities. This is a study that's not yet published of 2,200 years ago in Yorkshire in northern England. And you see extreme matrilineality. This person's mom, his mom, his mom, his mom, always going back into the past. And so you see this too in these tombs.
So the final thing I wanted to talk about was how we're beginning to be able to learn about biology. The hope with ancient DNA was always to learn as much about biology as about history. But really, you can learn a tremendous amount from a few genomes, because a single person — a genome is not a single person. It's a multitude of people, because it contains within it your parents and your parents' parents and your parents' parents' parents and so on. So you can get exquisite precision on the ancestry of each person based on just a single genome.
But to learn about biology, you need to track the frequency changes over time of a variant. And then you need large sample size, which we now begin to have. So ancient DNA studies between 2014 and 2015 really found only a handful of mutations that changed too rapidly in the time series data produced by ancient DNA for that to be due to chance. Here's a study I was involved in, led by Iain Mathieson, who was a postdoc with me at the time, that did this in 2015, and we found 12 locations in the genome that must have changed due to natural selection.
But in the next nine years that only increased to 20, and it made us think that maybe there was actually not a lot to be found. And there's been really an orthodox view in population genetics, which is that rapid adaptation has been rare in the last couple of hundred thousand years, there's no strong selective sweeps, and that we're actually not changing very much under the pressure of natural selection.
So what we've done is we've deployed a few methodologies to increase power and reduce false signals. One of them is we're not explicitly modeling the demographic history. We're letting the data just teach us the correlation structure of the data. The second is that we've increased the sample size by a huge amount. So in the last year we've reported data that's sort of close to double the amount of data in the literature. So we're releasing data pre-publication, or we've gotten permission to release data in this way, in order to increase the sample size. And then we're using statistical imputation to effectively double the sample size further.
So when you do this, you find hundreds and hundreds of positions that are clearly under natural selection, and all sorts of tests and second lines of evidence and third lines of evidence makes it clear that these are operating under the pressure of natural selection, and people are already using this in all sorts of ways. So how to recognize these results with the lack of evidence of selective sweeps in the previous genetic data? Well, one answer is that we're in a period of intensified evolution, at least in this part of the world, which is Europe.
There seems to have been intense evolution the last 8,000 years, maybe because people were moving to new cultural and environmental conditions with an especial intensity and over-representation compared to the background that one would expect for variants affecting metabolism and immune traits. And there's not only that, but there's been a further acceleration after 5,000 years ago. So you can measure the intensity before 5,000 years ago and after 5,000 years ago, and there's an acceleration.
But the second answer is not just that it's accelerated, but it's fluctuating. So if you look at a plot of the age of the genetic variant when it arose in the population versus what it would be if it had been selected at that intensity for all time, you see many positions on the y equals x axis. And that's what you expect if it's a constant selective sweep. But there's many that are on the right there that are circled that are much older in truth than you would expect. And this must mean because selection has increased over time, or even reversed directions, as shown by the red dots.
I'm just going to have a few more slides, and I think I have a few more minutes. And you can see selection on particular important mutations, like for example celiac disease resistance — multiple of them have rocketed up in — not resistance, but susceptibility — have rocketed up in frequency in the last 4,000 years. They cause a disease, but maybe like sickle cell, they also protect against something.
Here's a tuberculosis severe risk factor which has reversed in direction, rocketing down 3,000 years ago before it jumped up in frequency — selection in both cases. So maybe this is exposure to tuberculosis for the first time, or maybe something else. You can also look at combinations of variants that affect complex traits. So for example, type 2 diabetes — we heard about autism before — there are many combinations of variants, and you can make a score for trying to predict the trait based on a combination of variants.
So there's several type 2 diabetes risk factors that have decreased very significantly in frequency over the 10,000 years. So this is fat mass, body fat mass. This has decreased by one standard deviation. And this is consistent with this thrifty gene hypothesis that maybe as people had more stable food availability, maybe there was less need to store fat.
Another thing is we see is that the genetic susceptibility predictor of schizophrenia has also decreased dramatically. There's clearly been selection against what today manifests itself as schizophrenia. This is not to say that there was more schizophrenia 5,000 years ago than there is today. It's quite possible that the environment today makes people more likely to get bipolar disease and schizophrenia. Bipolar disease is also affected — negatively selected in this time.
And then you also — there's genetic predictors of, believe it or not, the number of years people attend school that are highly statistically significant. And this has clearly been under positive selection. And we've tested this all sorts of ways. And this doesn't mean that people were going to school less often in the past than today, because there was no school. So there it was clearly selection for something else that today correlates to this, and for example it correlates to very different things, such as for example the age at which a woman chooses to have her first child — a behavioral trait, not a fundamental fertility trait. So it has something to do with behavior.
So we're finding through this increase of number of samples — what was found before this year were these outliers, these extreme unusual cases of natural selection. But we now have enough sample size to find these many signals because we have enough power, because we've driven the standard errors around the estimates of selection. And that's clearly where the rich seam of information about biological adaptation is. And it's very cheap to generate this data. There's millions of skeletons that can be analyzed. So in principle, one could do this in many contexts.
There's a huge amount of data now from Europe. It's about 80% of the data. But there's huge gaps also in many other parts of the world, and availability to study these places to compare and contrast, and it seems an important direction both for studies of history and biology.
Current funding landscape in the US is a — whoops — is a challenge for sustained progress. There's a lot of things going on in our laboratory right now, but we're needing to cut staff and samples a lot right now because of these issues, and it's really a big compounding existing problems. In Europe there's dozens of ERC grants, Max Planck departments. In the USA there's never been a federal grant that supported this type of research.
In our own work, a pillar of our own support has been Howard Hughes Medical Institute support since 2013. But most of our funding came from the John Templeton Foundation and the Paul G. Allen Frontiers Group, and also the NIH. All that's about to stop or now stopped. So I think generous philanthropic support, such as could come from the Simons Foundation and other places, could make a big difference to us, labs like mine, and science in this and other areas. Thanks. [Applause]
