Statistics used to infer inter-breeding between humans and Neanderthals are strongly predicted by flanking sequence heterozygosity.

A large and rapidly expanding literature has grown out of the observation that humans carry a genetic legacy reflecting ancient inter-breeding with archaic hominins such as Neanderthals and Denisovans. However, a recent study suggests that a commonly used statistic used to assess legacy size, D, is driven mainly by heterozygous sites in Africa acting to increase divergence from our common ancestor rather than introgressed fragments outside Africa reducing divergence. To test this new model, I analysed how D is influenced by heterozygosity within a kilobase of each putative introgressed base. I find that flanking heterozygosity is a potent predictor of D, with introgression always being inferred as having occurred into the population with lower heterozygosity. This pattern cannot be driven by any introgressed fragments themselves, which simulations show create the exact converse pattern, but instead appears to be generated by heterozygosity acting to drive increased divergence from the ancestral sequence. This new model explains why introgression of haploid or semi-haploid regions is essentially lacking and why introgression is often inferred around immune genes and other regions under strong selection. More generally, these results raise the possibility that reported legacies are largely an artefact arising out of the false assumption that mutation rate is constant.