Amazonians may have Andaman-like admixture


Recently had a discussion with a reporter at a major publication about genetic genealogy, and how genomics and ancient DNA has changed everything about what we know about the human. Though I did put in the caveat that it seems the New World has a mildly simpler history that aligns with what we’d somewhat expected or seen. I was wrong. It turns out that indigenous people of the Amazon have a few percent of ancestry derived from a population with the closest affinities to those of Australasia.

Screenshot from 2015-07-21 10:46:26The plot to the left roughly shows groups which share haplotypes when comparing the Mixe people of Mexico to other Amerindians. The stronger red shading shows an affinity between pairwise groups due to the putative admixture event. What you can see that various Australasian groups, as diverse as the Andaman Islanders and and Papuans show elevated signals of affinity with particular groups from the Amazon in comparison to the Mixe.

The model is outlined in the figure at the top. What you can see is that a there was admixture into one of the first groups which arrived in the Americas. This group already had ancient Siberian ancestry. It then seems to have absorbed some ancestry from another group with affinities to the peoples of Australia. Because the dominant ancestral component of the hybrid group was similar to the sister American lineage a wide range of ancestral fractions from this group into the peoples of the Amazon are compatible with the statistics yielded by the data. Up to 85 percent, or as low as 2%. The lower the fraction of admixed “First Americans” into the Amazonian groups in question, the higher the Australian element in that group, as that ancestral element only spans a 1-2% range. In other words, if the admixed First American group contributed only 2% to the modern Amazonians, it would be in large part descended from the Andaman-like group.

This detail matters because of what is brought up in the discussion of the paper, and has been mooted elsewhere: a lot of the older skeletons from this region of South America look different. One model, which the authors are skeptical of, is that admixture happened in the New World. That is, and outrider group of First Americans absorbed and older population which arrived earlier to the region. But Kennewick Man tells us that morphology is only a rough guide. The Anzik Clovis result did not yield any evidence of Andaman-like admixture. That means either there was structure coming into the New World. Or, the admixture happened here.

From the patterns in the genome the admixture is old. But the confidence interval is big. It is not detectable by those methods sensitive to recent admixtures. It is older than 4,000 years. But, it is clearly post-Neandertal admixture. So younger than 50,000 years. In fact, the admixture graph at the top of this post suggests that this admixture post-dates that with the ancient Siberians. That means it is highly likely it probably dates to no older than ~20,000 years, since you need some time for the various Eurasian groups to actually diverge from about 40,000 years ago onward.
dstatThe authors used a variety of methods to test this affinity that they detected. The ancestry fractions were low, so they wanted to see if the result was robust. It sees to be. I won’t get into the details, but I want to post the D-statistic table from the extended data. It uses a slightly different method the map above. Basically they’re testing explicit trees, and showing deviations from random drift along the independent paths which are compatible with gene flow across the tips of the tree. Admixture.

It strikes me that D-statistics show a lot more affinity with South Asians broadly. Why weren’t the signals as strong as in the Chromopainter analysis above? I have no idea, but perhaps it might have to do with the fact that South and Southeast Asians are themselves admixed between the Andaman-like group, and other populations (West and Northeast Eurasians respectively). This may have made higher to detect the haplotypes in question. Additionally, the Ami, an indigenous people of Taiwan also show up on this list. What this implies is that a broad constellation of eastern peoples who diverged from the ancestors of Northeast Eurasians were widely present in the past. In South and Southeast Asia descendants of these people were probably dominant down deep into the Holocene. The elevated signatures in groups like the Andaman Islanders and Papuans may be due to the fact that these groups are relatively pristine.

gr3In 2011 the Reich group published a paper which actually suggested that there were multiple waves of old Southeast Eurasians with the Andaman Islanders being remnants of a group which contributed a secondary signal of admixture into the peoples of Oceania. This was published at about the same time as a whole genome analysis which suggested that Oceanians were descended from an earlier migration out of Africa, and not admixed. There are other groups which are supporting multiple out of Africa events now. The plot is getting thick and complicated.

I only bring this up because much of the current work uses copious data to test explicit models. The authors are constraining the sample space of models, and if you select between ten models, it may still be that the best fit is not a really good fit to the “real” model of what actually happened. One reason ancient DNA has been revolutionary is that it has forced researchers to consider models that they would have otherwise thought ludicrous. There were long suggestions for example of common ancestry between Amerindians and Europeans…but these ideas were often discarded as implausible. There are all sorts of details in relation to modeling assumptions which give valid results which are nevertheless inaccurate.

It is turning out that reality is crazier than our imaginations. Hold tight.

CiteGenetic evidence for two founding populations of the Americas, Skoglund et. al., Nature (2015) doi:10.1038/nature14895

Posted in Uncategorized

Comments are closed.