There’s an update in the debate about Neanderthal introgression into AMH in Eurasia. It’s based on some tests and simulations rather than in new ancient data, so it’s only interesting for the more inclined to look at stats, etc… Still, coming from the Svante Pääbo team and having to do with an important part of our deep prehistory I’ll post some quick thoughts about it.
The limits of long-term selection against Neandertal introgression – Petr et al 2018 – Preprint. From the abstract (emphasis mine):
Several studies have suggested that introgressed Neandertal DNA was subjected to negative selection in modern humans due to deleterious alleles that had accumulated in the Neandertals after they split from the modern human lineage. A striking observation in support of this is an apparent monotonic decline in Neandertal ancestry observed in modern humans in Europe over the past 45 thousand years. Here we show that this apparent decline is an artifact caused by gene flow between West Eurasians and Africans, which is not taken into account by statistics previously used to estimate Neandertal ancestry.
The problem they find with earlier estimates based on f4 ratios was the assumption that the relationship between Africans and West Eurasians had remained constant over time. However they found this isn’t the case:
To test for gene-flow between Africans and West Eurasians over time, we calculated D statistics in the form of D(Ust’-Ishim, X; African population, Chimp), which tests for changes in affinity (i.e. the number of shared derived alleles) between a series of West Eurasians (X) and Africans with respect to Ust’-Ishim – a 45,000 year-old Eurasian individual who is the oldest modern human genome in this dataset. In the absence of gene-flow between a West Eurasian lineage X and Africans in this time-frame, the value of this D statistic will not be significantly different from 0 for any of the West Eurasians regardless of their age. However, in violation of the assumption of the indirect f4-ratio, we find that starting from as early as 20 thousand years ago (kya) this D statistic becomes increasingly negative, consistent with gene-flow between West Eurasians and Africans during this time (Fig. 2). To rule out potential technical issues, we repeated this analysis substituting Oceanians for Africans and find no changes in affinity over time (Fig. 2).
With the new data from the Vindija Cave (Croatia) Neanderthal, they instead performed an f4 ratio using 2 Neanderthals to test for admixture in West Eurasians above what Africans could have, in the form f4(Altai, Chimp; X, African) / f4(Altai, Chimp; Vindija, African), where X are West Eurasian populations from 45 thousand years ago to present day. With this “direct” f4 ratio they find no decrease in Neanderthal admixture over time.
They confirm that West Eurasians have admixture from a population that split before East and West Eurasians did (the hypothetical Basal Eurasian population) by using a stat in the form f4(West Eurasian W, East Asian X; Ust’-Ishim, Chimp), which is significantly negative. However, they reject the idea that this Basal Eurasian population had no Neanderthal admixture, since populations with close to 50% Basal Eurasian admixture don’t have significantly less neanderthal admixture than those without Basal Eurasian admixture (according to their new above findings).
The paper goes on with some other considerations and simulations about the negative selection against deleterious alleles from Neanderthals in AMH, noting that this would only happen significantly enough during the first generations after the admixture event, but not significantly so after that initial period.
And now for some thought about it:
First they test D(Ust’-Ishim, X; African population, Chimp) with X being a set of West Eurasians and find that starting 20 kya it becomes increasingly negative, as seen in Fig 2. This has 2 possible interpretations:
- There’s gene flow between Africans and West Eurasians starting around 20 kya (their hypothesis)
- There’s decrease in Neanderthal admixture in West Eurasians starting 20 kya (the previous hypothesis)
The second interpretation is based on the idea that lower Neanderthal admixture would indeed make West Eurasians (starting 20 kya) being closer to Africans than Ust-Ishim (since Neanderthal admixture would push Ust-Ishim further away from Africans). But I’m not sure if they’ve considered the second option since they didn’t really provide the relevant stats to discard it. They do provide a second stat in the form D(Ust’-Ishim, X; Oceanian population, Chimp), which unlike with Africans does not become negative with younger samples. But this stats are difficult to interpret by themselves. Besides, it would be still based on another assumption: that the relationship between Oceanians and West Eurasians is constant over time. And that’s not clear, though it would be easy to test. Is, for example, D(Kostenki14, Loschbour; Oceanian, Chimp) not significantly different from 0? As far as I remember that would be negative, which would require an explanation as to why this doesn’t show in the other stats with Oceanians.
So how to prove that the first interpretation is correct and not the second one? That seems rather straight forward. If there’s West Eurasian admixture into Africans, then stats in the form D(West_Eurasian, East_Eurasian; African, Chimp) should be significantly positive. Is it the case? Again, as far as I remember, that’s not the case with West and Central Africans, but the paper does not provide the stats to check it.
About their results when using f4(Altai, Chimp; X, African) / f4(Altai, Chimp; Vindija, African), that’s a good test, but not so straight forward to understand well. How about confirming the pattern with a different, but theoretically equivalent, approach? Something like D(X, Y; Neanderthal, Chimp), where X is a West Eurasian sample older than 20 ky and Y a West Eurasian sample younger than 20 ky. They should be ~0, and maybe they are, though I’d like to see it tested.
Then about the hypothetical Basal Eurasian population and its lack or not of Neanderthal admixture this is rather difficult to test, since we don’t have any simple of that hypothetical population. The bigger problem is that we cannot even have a good estimate of how much Basal Eurasian admixture is present in modern (and some ancient) West Eurasians. Estimating that Early European Farmers had ~44% Basal Eurasian admixture is a mere hypothesis and not very robust (further attempts for estimates in the later Lazaridis et al 2016 paper show how frustrating it can be to try to estimate that).
I welcome any attempt to figure out all these details about early Out of Africa events, archaic admixture, Basal Eurasian origin, etc… but I can’t see that this paper provides new and solid insights into it. I’ve seen many attempts and lots of formal stats trying to solve these questions and I can say that a definitive answer won’t come until we get relevant ancient DNA to sort it out. Nevertheless, I still prefer the explanation that Upper Paleolithic samples have more archaic admixture than younger ones. It looks more compatible with all the tests I’ve seen over time. As for Basal Eurasian, I still find it difficult to explain that it represents a population 100% from the main Out of Africa event. It’s inconsistent with the data I’ve seen so far. A better explanation is a smaller amount of a more divergent branch (African, I’d guess), which would make the dilution of Neanderthal admixture very small too and hard to assess with accuracy. All in the realm of speculation, in any case.