Decomposing the admixture statistic, D, suggests a negligible contribution due to archaic introgression into humans.

Abstract

It is widely accepted that non-African humans carry a few percent of Neanderthal DNA due to historical inter-breeding. However, methods used to infer a legacy all assume that mutation rate is constant and that back-mutations can be ignored. Here I decompose the widely used admixture statistic, D, in a way that allows the overall signal to be apportioned to different classes of contributing site. I explore three main characteristics: whether the putative Neanderthal allele is likely derived or ancestral; whether an allele is fixed in one of the two human populations; and the type of mutation that created the polymorphism, defined by the base that mutated and immediately flanking bases. The entire signal used to infer introgression can be attributed to a subset of sites where the putative Neanderthal base is common in Africans and fixed in non-Africans. Moreover, the four triplets containing highly mutable CpG motifs alone contribute 29%. In contrast, sites expected to dominate the signal if introgression has occurred, where the putative Neanderthal allele is absent from Africa and rare outside Africa, contribute negligibly. Together, these observations show that D does not capture a signal due to introgression but instead they support an alternative model in which a higher mutation rate in Africa drives increased divergence from the ancestral state.