Figure 5. Selection at a single locus impacts genome-wide ancestry proportion.

(A) Inferred (dark gray), simulated (white), and observed (red) mean of global ancestry in Santiago over time. The dark gray histogram plots the posterior distribution for initial $(g=1)$ West African ancestry contribution inferred using approximate Bayesian computation (ABC) (median, 0.690); the prior distribution [${\displaystyle m\sim U(0.1,0.9)}$] is in light gray. The red line plots the mean global ancestry estimated by admixture from modern genetic data from Santiago, 0.737. The observed global ancestry is higher than most values of the initial contributions inferred in dark gray. The white histogram plots the distribution of West African global ancestry proportion calculated after 20 generations in populations simulated with selection coefficients and initial ancestries drawn from the ABC-inferred values (median, 0.723). The global ancestry calculated after 20 generations of simulated selection (white) more closely matches that observed from Santiago genetic data (red line). (B) West African mean global ancestry proportion calculated for 500 simulated populations after 20 generations under varying single-locus selection coefficients, ${\displaystyle s}$. We simulated whole autosomes, setting the initial West African ancestry contribution to 0.65. Black circles indicate mean ancestry on chromosome 1 alone. Gray circles indicate mean ancestry on the other autosomes (2–22). The increase in ancestry with selection for gray circles demonstrates that selection impacts global ancestry beyond the local effects of the chromosome under selection.

Figure 5—figure supplement 1. Effect of selection on global ancestry across simulation methods.

Pink circles indicate West African mean global ancestry after 20 generations versus selection coefficient for whole autosome (22 chromosome) simulations, using a uniform recombination rate within each chromosome. Green triangles represent mean weighted ancestry for chromosome 1 and chromosome 2, with chromosome 2 representing the 92% of the genome that segregates independently from chromosome 1, using a human genetic map for recombination rates. We performed 500 simulations for each model, and all simulations started with West African ancestry contribution of 0.65. The estimated slope and intercept for the two methods of simulating global ancestry are highly similar. ANCOVA results suggest there is a significant effect of selection coefficient on global ancestry: F(1,997) = 1.0519 × 10⁴, p < 2 × 10⁻⁶, but there is no significant effect of recombination rate and simulation model on global ancestry estimate after controlling for selection coefficient: F(1,997) = 1.6350 × 10⁻¹, p = 0.686.