Sunday, October 4, 2015
Yamnaya's exotic ancestry: The Kartvelian connection
I've made a discovery. The Near Eastern-related ancestors of the Yamnaya steppe pastoralists were also the ancestors of present-day Georgian Mingrelians, or their very close relatives, and in all likelihood speakers of Kartvelian, which has a long history in the Caucasus. Here's a nice map from Wikipedia and a pic of some Mingrelians. Check out the impressive headware.
TreeMix is very specific and precise about this. In my analyses, based on a couple of different methods, the Mingrelians are the only population chosen as a source for the Near Eastern-related ancestry in the Yamnaya.
Keep in mind, this is an unsupervised test and the algorithm has an infinite number of choices, because migration edges can run from any part of the tree, and yet it chooses the Mingrelians. By the way, if anyone's wondering, I did also try the Bronze Age Armenians, to no avail.
This outcome is also more or less reproducible with more complex topologies that include samples from Central Asia. In the graph below the Georgian Mingrelians form a clade with the Near Eastern-related ancestry of the Yamnaya. It'd be interesting to see if other Georgian groups, like the Svans, do even better, if that's actually possible, but they're not available at the moment.
I actually came up with basically the same result earlier this year using qpAdm (see here). But at the time I was skeptical of its usefulness because qpAdm only offers a supervised test, so picking Georgians as a reference population and getting a good statistical fit doesn't mean as much as a reproducible unsupervised migration edge.
Now, judging by their ADMIXTURE results, these Georgian Mingrelians do carry some Early European farmer-related ancestry, which is missing in the Yamnaya (see here). Therefore, it's likely that ancient samples from the west or northwest Caucasus will prove to be even better proxies for the Near Eastern-related ancestry in the Yamnaya.
The samples used to produce the above TreeMix graphs are listed here. They're sourced from the Allentoft et al., Haak et al., and Lazaridis et al. datasets. I limited the markers to ~65K transversion (high confidence) SNPs that overlap between these datasets.
Caucasus hunter-gatherers (CHG) and the Indo-European question
'Fourth strand' of European ancestry originated with (Caucasus) hunter-gatherers isolated by Ice Age
Mixed marriages on the early Eneolithic steppe