As some of you know the supplement on the Cardial ancient genome presents some really strange results. Look at the gene flow edges around the Mbuti. I have used TreeMix a fair amount, and I’ve seen some weird things of which I have not spoken. I’m sure others have as well. My assumption when I see something this peculiar is that it is an artifact of ascertainment bias or TreeMix just being TreeMix.
After the welter of novel analytic methods and ancient DNA I think perhaps we need to reconsider dismissals. Usually when you’re using TreeMix you’re looking for only a constrained set of the hypotheses. You pretty much ignore or shove aside “results” which bubble out of the woodwork. But sometimes this marginalia can take center sage. When the original TreeMix paper came out it had a figure which indicated gene flow from the base of the Amerindian lineage to Russians. This was very peculiar. We now know what was happening. It was the signal of the admixture of the Ancient North Eurasians, represented by the Mal’ta boy. Similarly, the evidence of Austro-Melanesian admixture in some Amazonian peoples has been there in the SNP-chip data for years. It was just overlooked. Why? Because it didn’t make sense.
The examples are pretty numerous now. What is going on in my opinion is that the method of TreeMix is searching for the nearest proxy for donor populations when those populations are no longer present. That produces bizarre results in some cases, as the topologies confound our expectations. But the problem isn’t in the topologies, it’s in our interpretative framework. A reticulated genetic graph filled with implicit positions which are representative of extinct lineages which nevertheless contributed genes to extant lineages is going to be harder to unpack than our previous understanding. But what needs to happen is we need to update our understanding….