Bias in estimators of archaic admixture

Abstract

This article evaluates bias in one class of methods used to estimate archaic admixture in modern humans. These methods study the pattern of allele sharing among modern and archaic genomes. They are sensitive to "ghost" admixture, which occurs when a population receives archaic DNA from sources not acknowledged by the statistical model. The effect of ghost admixture depends on two factors: branch-length bias and population-size bias. Branch-length bias occurs because a given amount of admixture has a larger effect if the two populations have been separated for a long time. Population-size bias occurs because differences in population size distort branch lengths in the gene genealogy. In the absence of ghost admixture, these effects are small. They become important, however, in the presence of ghost admixture. Estimators differ in the pattern of response. Increasing a given parameter may inflate one estimator but deflate another. For this reason, comparisons among estimators are informative. Using such comparisons, this article supports previous findings that the archaic population was small and that Europeans received little gene flow from archaic populations other than Neanderthals. It also identifies an inconsistency in estimates of archaic admixture into Melanesia.

Evolutionary history of Tibetans inferred from whole-genome sequencing

“…However, we observed no significant difference in D between Tibetans and Han Chinese (D = -0.0007, 95% CI -0.0027 to 0.0021), demonstrating that the genome-wide level of Denisovan admixture does not differ substantially between Tibetans and Han Chinese, and is consistent with a previous report[47]. Using the Q statistic of Rogers and Bohlender[48], we estimated that the proportion of Denisovan admixture in Tibetans is 0.4% (95% CI: 0.2% to 0.6%, see Methods and S4 Fig). Briefly, Q is an f 3 ratio estimator of admixture[48].…”

“…Using the Q statistic of Rogers and Bohlender[48], we estimated that the proportion of Denisovan admixture in Tibetans is 0.4% (95% CI: 0.2% to 0.6%, see Methods and S4 Fig). Briefly, Q is an f 3 ratio estimator of admixture[48]. Given a sample from three populations X, Y and N, so that X and Y are more recently diverged from each other than either is from N, the calculation of Q uses a ratio of expectations to calculate the total admixture from N into Y, in a similar way as other f 3 -ratio estimators[46].…”

The apportionment of human diversity revisited

“…Previous studies have demonstrated that nonsub-Saharan African populations may have a small amount of Neandertal genetic ancestry Reich et al, 2010;Pr€ ufer et al, 2014). While the precise timing and amount of admixture continues to be refined (Skoglund and Jakobsson, 2011;Pr€ ufer et al, 2014;Rogers and Bohlender, 2015), the systematic pattern of lack of fit that we identified in this study is consistent with Neandertal introgression in the first modern population to leave sub-Saharan Africa.…”

Early history of Neanderthals and Denisovans

“…This view may be an artifact of ascertainment bias (17) or of the biases documented by Rogers and Bohlender (10). On the other hand, the East Asian excess may be real, but hidden by the broad confidence intervals surrounding our estimates of mN .…”

“…Their "ABBA-BABA" statistics infer admixture from the frequency with which derived alleles are shared by pairs of samples. As we have shown (10), these estimators have large biases when populations receive gene flow from more than one source. The magnitudes of these biases depend on the sizes and separation times of ancestral populations.…”