Skip to main content

Exploring qpAdm / qpGraph models

I think this particular mini project could have—out of every other case study idea on offer—the most impact on you your research in ancient DNA genomics.

Both qpAdm and qpGraph techniques form a core component of nearly every single ancient DNA paper. Yet, a lot of research work has been poured (one paper, another paper, yet another paper) into delineating in which situations qpAdm works and — much more often — when it either fails to work or when it’s often completely misinterpreted.

If you work with qpAdm, you’re aware of many assumptions and recommendations that are made, but those are rarely backed by rigorous modeling and simulation.

In this case study, I would encourage you to take some qpAdm / qpGraph models you have inferred in the past (or some foundational models inferred in the literature relevant for your work), program them in the form of a parametrized slendr model, and experiment with how robust the qpAdm conclusions are to some of the assumptions made.

Hint: You already know that slendr allows exporting data in the form of EIGENSTRAT files via the function ts_eigenstrat() (as introduced in recipes in this chapter) which means that connecting slendr simulation scripts with the ADMIXTOOLS2 software in R should be quite straightforward.