New work from the lab! We present a novel method for inferring the geographic locations of shared genetic ancestors and use it to infer the geographic history of human genetic ancestry biorxiv.org
1/n This work is led by postdoc in my lab Mike Grundler (not on twitter) and is in collaboration with Jonathan Terhorst (also not on twitter).
2/n The method works by using Sankoff parsimony on the local genetic genealogies to infer minimum migration cost surfaces for each ancestral node (conceptual illustration below). The algorithm is fast; we can analyze genomic datasets w/ 100s-1000s of samples.
3/n We tested it using simulations under a variety of scenarios and dispersal kernel shapes, and found that we can accurately recover ancestor locations, even relatively far in the past
4/n We then apply it to a large sample of humans to infer the geographic history of human genetic ancestry. The plot below shows a visualization (a "tanglegram") of all the branches that connect all the modern day samples to all of their ancestors through space and time.
5/n Note that this is not the same as the geographic history of human dispersal! We can only learn about *shared* genetic ancestors from the sequenced *sample*.
6/n Also, important to note that the minimum cost migration surfaces become a lot flatter for ancestors deeper in the past because there's less information about exactly where they lived
7/n Despite these caveats, we think this is a useful approach. See below for the tanglegram for a single individual (in white): lots of recent ancestors ALL OVER; median ancestor location tracks out of Africa
8/n We think the method offers a big step forward for how to think about ancestry (particularly in humans): best defined with *explicit* reference to a point in space AND time.
9/n E.g., how much of your genome did you inherit from ancestors that lived inside some region X years ago. Using this definition, ancestry is not static - it changes through time as your ancestors (carrying the bits of genome that will end up in you) moved around.
10/10 Really excited for feedback on this - please reach out w/ thoughts/suggestions!
Loading suggestions...