Substack cometh, and lo it is good. (Pricing)

Ancient DNA Selection Paper Just Dropped!

1,000 ancient genomes uncover 10,000 years of natural selection in Europe:

Ancient DNA has revolutionized our understanding of human population history. However, its potential to examine how rapid cultural evolution to new lifestyles may have driven biological adaptation has not been met, largely due to limited sample sizes. We assembled genome-wide data from 1,291 individuals from Europe over 10,000 years, providing a dataset that is large enough to resolve the timing of selection into the Neolithic, Bronze Age, and Historical periods. We identified 25 genetic loci with rapid changes in frequency during these periods, a majority of which were previously undetected. Signals specific to the Neolithic transition are associated with body weight, diet, and lipid metabolism-related phenotypes. They also include immune phenotypes, most notably a locus that confers immunity to Salmonella infection at a time when ancient Salmonella genomes have been shown to adapt to human hosts, thus providing a possible example of human-pathogen co-evolution. In the Bronze Age, selection signals are enriched near genes involved in pigmentation and immune-related traits, including at a key human protein interactor of SARS-CoV-2. Only in the Historical period do the selection candidates we detect largely mirror previously-reported signals, highlighting how the statistical power of previous studies was limited to the last few millennia. The Historical period also has multiple signals associated with vitamin D binding, providing evidence that lactase persistence may have been part of an oligogenic adaptation for efficient calcium uptake and challenging the theory that its adaptive value lies only in facilitating caloric supplementation during times of scarcity. Finally, we detect selection on complex traits in all three periods, including selection favoring variants that reduce body weight in the Neolithic. In the Historical period, we detect selection favoring variants that increase risk for cardiovascular disease plausibly reflecting selection for a more active inflammatory response that would have been adaptive in the face of increased infectious disease exposure. Our results provide an evolutionary rationale for the high prevalence of these deadly diseases in modern societies today and highlight the unique power of ancient DNA in elucidating biological change that accompanied the profound cultural transformations of recent human history.

Sometimes statistics confirm what you can already intuitively discern. I’m an avid reader of the ancient DNA and pigmentation literature. It was immediately obvious to me that the Bronze Age Northern European samples were darker than modern populations in that region. There are massive sample sizes for modern Europeans, and it jumped out at me that the derived SNP at SLC45A2 was at a lower frequency two to four thousand years ago than it is today. This paper confirms that selection not only occurred during the Bronze Age, but has been happening in the historical period…

Note: The backstory on why I had my suspicions about this goes back to 2009. When the first Reich lab India paper came out I was pretty skeptical that the “Ancestral North Indians” were so similar to Northern Europeans. My rationale was straightforward: we should see way more blonde and blue-eyed people in Northern India and Pakistan. Nick Patterson pointed out the relevance of natural selection. Perhaps the Northern Indians got darker? There is a bit of evidence of this for some loci, but the Narasimhan et al. paper had a bunch of Sintashta, who are probably the best ancestral population proxy for the early Indo-Aryans, and they were clearly just not as lightly pigmented as related people on the Baltic today. The Sintashta were very similar to their relatives in Poland genetically, but they diverged from their Northern European relatives 4,500 years ago as the Fatyanovo population migrated eastward, and were probably in India when selection was changing the allele frequencies around SLC45A2.

19 thoughts on “Ancient DNA Selection Paper Just Dropped!

  1. Any info on change in loci affecting psychological traits like schizophrenia, ADHD, autism .etc

  2. Probably we could say now that thus far geneticists (and their funders) gave too mach importance to superficial traits (for instance the colour, thus WHGs were “black”) and not to ideology and “history” that is what actually divides men. And also to do that had its purposes.

  3. The Extended Figure 3 is very informative on SLC45A2. The figures there show the frequency of the selected variants they detect in all groups, and the expected frequency based on an admixture model of the previous groups (as a dashed line).

    The Steppe sample (Yamnaya and Afanasievo only) and the European Neolithic (dated to an average of around 3,400 BCE with the majority of samples from Central and Western Europe) are exactly the same. The Bronze Age (dated to an average of around 2,000 BCE largely from the Bell Beaker cultures of Czech Republic, Great Britain, Germany, and Slovakia) is quite different and increased. The Beaker period can be inferred to have a quite rapid upswing.

    As an aside on this, one plausible idea that people raised in comments on forums and blogs was that, although the steppe people didn’t have modern day frequencies of SLC45A2, it was higher and when admixing, they associated it with social class and ethnic identity and pushed it higher. This seems to throw a bit of doubt on that, potentially, although it may be true beyond the coarser scale of the resolution here.

    The derived variant at HERC/OCA2 linked to blue eyes also shows exactly no differentiation between their steppe and European farmer (Middle Neolithic) group either.

    Having just skimmed it, I would caution guess that it’s hard to be too precise from this; for example “Historical Period” is “Historical era individuals from the Roman and Late Antique periods, primarily from Britain dated to a mean age of ~2,000 BP”.

    So when they talk about selection in the historical period for LP or SLC45A2, maybe they are talking about the Iron Age and Late Bronze Age really. Some care with the labels they’ve used which are aggregating various things and skewed to particular times and places by availability. Like, I wouldn’t interpret this as “Europeans lighter in 1900 CE than 200 BCE”, though it clearly does support Europeans being lighter in 1900 CE than 2000 BCE.

  4. (I’d probably label the Historical Period, “Early Historical Period” here if I had to, due to it being an average around 2000 BP/0CE, although like all labels this has problems as history is earlier in some places etc…)

  5. The detection of selection on FADS 1/2 is maybe kind of affected here by the choice to use the binned populations rather than a continuous time approach. In this paper they don’t detect any selection between European Middle Neolithic and Bronze Age, while they do between BA and Historical, and European Middle Neolithic and Anatolian Neolithic.

    Mathieson’s paper estimated there to be a continuous time selection on this at an even, constant rate:

    So maybe this falls out of this approach. Here they claim that “we see that the major signal of selection at this locus is focused on the most recent epoch” (e.g. Bronze Age to Early Historical), which has some conflict with an even rate of selection.

    As an aside it’s kind of unusual that the steppe group has the same frequency of this variant as the European Middle Neolithic. It’s proposed to be an adaptation to more cereal based/agricultural diet, and the theory is that steppe group’s recent ancestors were all hunter-fishers and early pastoralists.

    On diet generally, kind of seems like they find selection in the neolithic to get less obese on a cereal based diet, which is lacking in the steppe pastoralist or mesolithic groups. Be interesting to see if this association still remains with steppe/neolithic ancestry today, like how there’s some evidence for contribution to the present day of the genetic height difference that probably existed at the time of 3000 BCE (which I would estimate would’ve been 1-2 inches).

  6. Razib: “My rationale was straightforward: we should see way more blonde and blue-eyed people in Northern India and Pakistan. Nick Patterson pointed out the relevance of natural selection. Perhaps the Northern Indians got darker? There is a bit of evidence of this for some loci, but the Narasimhan et al. paper had a bunch of Sintashta, who are probably the best ancestral population proxy for the early Indo-Aryans, and they were clearly just not as lightly pigmented as related people on the Baltic today.”

    Kind of looks to me like its a bit of both, although this is from memory. The selection after Yamnaya in Sintashta and related samples looks rapid, like in Beaker. And at the same time samples in Central Asia who have this ancestry have a decent frequency of the traits (as well as isolated Kalash who are likely to be isolated since formation of ANI?).

    But it was likely not near to the historical period (0CE) frequency by 1700-1200 BCE (the latter LBA-EIA part being where IMO its more likely steppe ancestry entered India and Iran).

    And Luca Pagani’s group found some indication of negative selection on SLC45A2 derived variant in India –“The local ancestry deconvolution in South Asians reveals opposite selective pressures on two pigmentation genes (SLC45A2 and SLC24A5) that are common or fixed in West Eurasians, suggesting post-admixture purifying and positive selection signals, respectively.”

    So it seems likely to be a mix of both?

  7. So when they talk about selection in the historical period for LP or SLC45A2, maybe they are talking about the Iron Age and Late Bronze Age really.

    the SDS method detects lightness selection in English into medieval times. the caution isn’t what you think it is; there isn’t power to detect really recent selection sweeps.

  8. Nah I disagree. Indians do have ancestry from real white people.

    The Indo-Iranians were white people.

    Sintashta had a derived allele frequency in SLC45A2 at 88% in your own analysis. This is near the modern European average of 95%.

    So yeah they were “white.” Can you tell me how I’m wrong?

  9. This idea you have that the Indo-Iranians split from northern Europeans early and so didn’t experience selection for light skin/hair is wrong.

    They didn’t truly leave Europe until Andronovo in 1800 BC.

    By this time, selective pressures had significantly increased the frequency of SLC45A2 in Europe.

    This is why Sintashta was 88% derived in SLC45A2 while Yamnaya was 40% derived.

    Yes, selection for lighter skin and hair continued in northern Europe after Indo-Iranians left but it had already made lots of progress by 1800 BC.

  10. Other papers sure could show some ongoing selection (on some level I think they do), but for this one, caution that their historical period sample does cluster around the dates that they’ve stated it does.

  11. In this case I actually don’t understand why they didn’t just extend the method with some contemporary people who are matched to the same locations as the Historical Era sample. Should be very available.

  12. It’s hard to agree with their idea that the SNPs upstream of DHCR7 were selected to better convert milk’s 7-dehydrocholesterol to vitamin D. Human ski contains lots of 7-dehydrocholesterol, and its concentration doesn’t seem to be limiting for vitamin D synthesis. When DHCR7 is disabled in Smith-Lemli-Opitz syndrome, then so much 7-dehydrocholesterol accumulates that it gets into bloodstream, but vitamin D production increases only ~1.5-fold
    The natural feedback loops meant to prevent vitamin D overproduction and toxicity probably work too effectively?
    In any case 7-DHC is effectively biosynthesized even without additional dietary intake.

  13. So yeah they were “white.” Can you tell me how I’m wrong?

    i have written how the sintashta would be called white but not nordic.

    if you leave any more comments misrepresenting my views again i will block and expurgate all your comments (if you reply to this comment, i will do so as well).

  14. Indians do have ancestry from real white people.

    “Real” white people? Are there fake white people in contrast? And what are the modern population groups that have ancestry from fake white people?

    Comrade, I condemn your obsession with whiteness and urge you to self-struggle!

  15. Reading more on DHCR7 … now I am less convinced. Perhaps it was affected by selection against vitamin D deficiency after all.
    Some SNPs in the general area (rs12803256, rs12798050) show quite decent effect sizes in vitamin D level GWAS, about 10% betas in Europeans and higher in East Asians. And papers show link with DHCR7 expression. G was the effect allele in the Europeans (increasing vitamin D), and it is uniquely enriched there, although, weirdly, it appears that a Korean paper reported an opposite effect direction? The overall impact is of course miniscule compared to the wide range of natural variability in vitamin D levels. But people argue that even small differences might have mattered in the past in population which teetered at the edge of severe vitamin D deficiency, because only sub-threshold values have health impacts. Like

  16. On phenotypes over time and Hirisplex, one bit of information, where this crosses over with the new Southern Arc papers.

    In the paper on “A genetic probe into the ancient and medieval history of Southern Europe and West Asia” (, they show data on the change in Hirisplex identified phenotype for West Eurasia over time, in Figure 6.

    Have a look at this one, as it shows a significant change in hair and skin and eye across West Eurasia as a whole over time.

    If we want to get more fine-grained, the supplement to that paper ( on page 18 in Table S5 on then breaks this down into more combinations, reporting data for populations with at least 5 individuals with enough data for all traits.

    If we’re less interested in the specific populations and more interested in the time trend, we can add to this the supplement from the main paper “The genetic history of the Southern Arc: A bridge between West Asia and Europe” again in table S5 (, where it reports for a lot more individuals who don’t meet the 5 individuals per population threshold.

    Once that’s aggregated, I separated all the data into increments of 500 year spans and 5 overall regions (North and Central Europe, South Europe, West Asia, South Asia and Central Asia).

    Here are some results for those, for the hair trait:

    The trends seem pretty apparent. There’s a decline in the black hair trait and emergence of more brown hair trait (and blond, with some more noise). Lazaridis hasn’t used all the data in the latest HO annos (because this paper has no doubt been in the works for a long time), and for North-Central Europe, recent papers could give a lot more resolution in the periods away from 2000 BCE. The presence of more Euro HG (with darker hair) in the early neolithic parts of the sequence for North-Central Europe makes the trend more dramatic there (as well as possibly stronger post 3000 BCE selection?).

    Also just for the Sintashta like samples since that kind of came up:

    It does seem like potentially these may be more likely to be blond haired than contemporary populations from the larger North-Central Europe set that I assigned (Switzerland, Czechia, Germany, Estonia, France, Britain, Hungary, Moldova, Netherlands, Poland, Russia, Austria, Ireland, Iceland, Lithuania, Latvia), albeit a little less likely than the medieval Europe NC-Europe set (largely Germany Medieval). Though maybe some structure there.

    The Sintashta-like samples are also driven here in a resurgence of the black hair trait and loss of the blond hair trait in the latest ~1000BCE period by these being entirely from the KAZ_Zevakinskiy_LBA subpopulation, who have quite a bit of North/East Asian ancestry. Could be that the Sintashta group was a bit “precociously” blond for its time and this trait got reduced in the Iranian offshoot populations by admixture and selection. (It would be kind of weird if this were actually true because as far as I know the idea that the proto-Indo-Iranians had relatively high frequencies of blond hair was based on some assumption, but it does seem like they are scoring relatively blond and little black hair on these Hirisplex results.) Though again this sample size is still relatively limited – the re-emergence of black hair here in the later population drawing on the more admixed Zevakinskiy_LBA is essentially driven by a single sample, etc.

    (Note, I had to assign dates to all these populations so it was quite a bit of business to check all this, but hopefully I got it right).

  17. @Matto

    At last some of my hardest opponents (above all who are 100% Italian like etrusco / old europe / etc i.e.
    Paolo Sizzi, and who are 50% like Richard Rocca, but the other 50% is from Iberia and the country is Argentine,
    thus not just far people and language) are coming back to my positions after 15 years of Jewish infatuation:

    Blogger Richard Rocca said…
    @Matt… It does not matter where RMPR437b plots on a PCA and it is not a good sample to use when looking for 1st generation migrants. It has a Bell Beaker R-L2 haplogroup and, below that, an FGC10543 subgroup that in modern
    samples is 99% found in the Italian peninsula. In ancient DNA, even with all the R-L2 Bell Beakers found to date, it has only been found in that Italic and another Iron Age Etruscan. It is obvious that this person had an Italic
    paternal line with an immigrant mother from the east.
    September 4, 2022 at 7:10 AM

  18. New preprint relevant to the same topic:

    “In admixed middle Neolithic individuals, we found excess Neolithic farmer ancestry at the pigmentation locus SLC24A5”

    This shows the same finding as the paper in Razib’s blogpost; SLC24A5 which was at high frequency in EEF farmers was subject to positive selection. E.g. admixture with WHG (Mesolithic) would be expected to reduce frequency but positive selection cancelled this out.

    However also note:

    In contrast to SLC24A5, the second high-effect pigmentation variant, in HERC, displays an excess of Mesolithic ancestry (+17.23%, |Z| = ~3.11).
    So the blue eyed depigmented iris variant shows an excess of Mesolithic ancestry, validating that there was selection for lighter eyes in the mixed Mesolithic+Anatolian group. E.g. MN farmers were lighter pigmented than the product of their admixing populations.

    On the other hand they also say: “the third high-effect pigmentation variant at SLC45A2, which arrived in Europe via later expansions from the steppe”, which we know is wrong, since the paper in Razib’s blogpost finds the same frequency of this variant in MN farmers and steppe populations…

    The finding of enriched WHG ancestry around HERC2 derived variant in MN farmers tends to validate the findings of Marnetto (, which some doubted, that showed among other things that more local WHG ancestry in present day Estonians in areas of the genome which influenc eye colour variation is correlated with lighter eyes.

Comments are closed.