Unexpected D-tour Ahead: Why the D-Statistic, applied to Humans, Measures Mutation Rate Variation not Neanderthal Introgression

It is widely accepted that humans interbred with Neanderthals and other extinct hominins, leaving a lasting genetic legacy. However, much of the supporting evidence was developed using the statistic D, which assumes, without testing, both that mutation rate is constant and that recurrent mutations are vanishingly rare. These assumptions together preclude an alternative explanation based on variation in mutation rates across human populations. Here we critically evaluate the assumptions underlying D and confirm that neither is valid. Over 40% of SNPs in dbSNP carry recurrent mutations. Theory indicates that D does not vary with mutation rate as long as the mutation rate does not vary between populations. In practice, D calculated separately for different sequence motifs varies greatly, implying strongly that mutation rates do vary between populations. We show that most, if not all, D-informative sites result from two mutations rather than the one mutation expected under the introgression hypothesis. Moreover, individual non-Africans carry a signal in more than five times as many genomic windows as can be accounted for by the 2% legacy they are thought to carry, indicating a signal that is radically more diffuse than expected. Remarkably, partitioning the data by whether the chimpanzee or Neanderthal allele is the major allele in humans reveals that the overall reported D-value of [~]5% actually comprises two opposing components: one with D [~] 30% and another with D [~] -25%. Tellingly, the positive component is produced by sites where the Neanderthal allele is the major allele, the exact opposite of what should be the case under introgression, where introgressed alleles should be rare. We show further that the entire D signal can be accounted for by sites where the Neanderthal allele is fixed outside Africa and the chimpanzee allele is rare inside Africa. Investigating potential mechanisms, we extend the published observation that the mutability of three-base combinations across human populations is influenced by flanking sequence heterozygosity to reveal how genomic regions that lost more heterozygosity out of Africa exhibit higher D-values. This correlation supports a model where loss of heterozygosity slowed the mutation rate, thereby reducing the divergence between Neanderthals and non-Africans. Across independent tests, our findings consistently indicate that the mutation rate variation hypothesis provides a more compelling explanation for the observed patterns in human-Neanderthal genetic relationships than the introgression hypothesis. We argue that the mutation rate variation hypothesis would help settle a number of conflicting patterns in the literature and, hence, that the concept of archaic introgression into humans and its implications for hominin-derived traits warrants reconsideration.