Fig 1.
Models of how genealogies are affected by selection at linked neutral sites.
The genealogies on the left represent species with a short split time such as human and chimpanzee. The genealogies on the right represent species with a long split time such as human and mouse. Red lines represent two lineages and their coalescent time. Blue lines represent two lineages and their coalescent time when there is selection at linked neutral sites in the ancestral population (abbreviated BGS). Yellow stars denote mutations accumulating on each of the two lineages after they split. Note that with the longer split time, the proportion of the genealogy attributed to the ancestral population decreases.
Fig 2.
Human-primate divergence is reduced at putatively neutral sites near selected sites.
(A) Neutral human-chimp divergence is negatively correlated with functional content. (B) Neutral human-orang divergence is negatively correlated with functional content. (C) Neutral human-chimp divergence is positively correlated with human recombination rate. (D) Neutral human-orang divergence is positively correlated with human recombination rate. Each point represents the mean divergence and functional content (A and B) or recombination rate (C and D) in 1% of the 100kb windows binned by functional content or recombination rate. Red lines indicate the loess curves fit to divergence and functional content (A and B) and divergence and recombination rate (C and D). The high variance of divergence at regions of low recombination rate is expected since the variance of divergence is inversely proportional to the recombination rate. Note that the last bin containing less than 1% of the windows was omitted from the plot. While the graph presents binned data, the correlations reported in the text are from the unbinned data.
Fig 3.
Human-rodent divergence is reduced at putatively neutral sites near selected sites.
(A) Neutral human-mouse divergence is negatively correlated with functional content. (B) Neutral human-rat divergence is negatively correlated with functional content. (C) Neutral human-mouse divergence is positively correlated with McVicker’s B-values. (D) Neutral human-rat divergence is positively correlated with McVicker’s B-values. Each point represents the mean divergence and functional content (A and B) or B-values (C and D) in 1% of the 100kb windows binned by functional content or B-values. Red lines indicate the loess curves fit to divergence and functional content (A and B) and divergence and B-values (C and D). Note that the last bin containing less than 1% of the windows was omitted from the plot. While the graph presents binned data, the correlations reported in the text are from the unbinned data.
Fig 4.
Models incorporating background selection can generate patterns of neutral divergence that recapitulate the empirical correlations.
(A) Models of background selection predict a positive correlation between neutral human-chimp divergence and human recombination. Because our model does not include biased gene conversion, the empirical correlation was calculated omitting AT to GC sequence differences. (B) Models of background selection predict a positive correlation between neutral human-mouse divergence and McVicker’s B-values. White histogram denotes 500 simulations not including background selection. Gray histogram denotes 500 simulations incorporating background selection (see text). Red line represents the correlation computed from empirical data. Thus, plausible levels of background selection can match the observed correlations while neutral simulations cannot.
Fig 5.
A two-locus model for the effect of background selection on divergence.
(A) The variance in divergence between two loci explained by background selection (BGS) as a function of the strength of background selection at the second locus (B2). (B) The expected proportion of divergence due to polymorphism in the ancestral population as a function of B2. (C) The variance in divergence between the two loci explained by polymorphism in the ancestral population as a function of B2. Different columns denote different mutation rates. Colored lines denote different ancestral population sizes (Na). Note that the variance in divergence attributable to background selection is greater than the expected proportion of divergence contributed by ancestral polymorphism.
Fig 6.
Relationship between divergence and functional content, human recombination, and McVicker’s B-values as a function of GERP score cutoff.
(A) Human-primate divergence versus functional content. (B) Human-primate divergence versus human recombination rate. (C) Human-rodent divergence versus functional content. (D) Human-rodent divergence versus McVicker’s B-values.