Table 1. Regression coefficient estimates for linear regression of F_ST on 2% quantile bins of B.

	AFR vs. EASN	AFR vs. EUR	AFR vs. SASN	EUR vs. SASN	EUR vs. EASN	SASN vs. EASN	Global
β0 ± SEM (p-value)	0.2044 ± 0.0039 (< 1e-04)	0.1716 ± 0.0031 (< 1e-04)	0.1596 ± 0.0029 (< 1e-04)	0.0455 ± 0.0011 (< 1e-04)	0.1216 ± 0.0029 (< 1e-04)	0.0903 ± 0.0023 (< 1e-04)	0.1322 ± 0.0019 (< 1e-04)
β1 ± SEM (p-value)	-0.0434 ± 0.0046 (< 1e-04)	-0.0358 ± 0.0037 (< 1e-04)	-0.0355 ± 0.0034 (< 1e-04)	-0.0098 ± 0.0013 (< 1e-04)	-0.0173 ± 0.0035 (< 1e-04)	-0.0261 ± 0.0027 (< 1e-04)	-0.0280 ± 0.0022 (< 1e-04)
r ± SEM	-0.8363 ± 0.0295	-0.7441 ± 0.0362	-0.7794 ± 0.0332	-0.3847 ± 0.0414	-0.6220 ± 0.0785	-0.5968 ± 0.0348	-0.1292 ± 0.0098

The first two rows give the regression coefficients for the linear model F_ST = β₀ + β₁B + ε, where B represents the mean background selection coefficient for the bin being tested and F_ST is the estimated F_ST for all population comparisons within a particular pair of continental groups (given in the column header). The final column, “Global”, gives the regression coefficients for the linear model applied to all pairwise population comparisons (150 total). The correlation coefficient, r, between B and F_ST for each comparison is shown in the bottom row. Standard errors of the mean (SEM) for β₀, β₁, and r were calculated from 1,000 bootstrap iterations (see Materials and Methods). P-values are derived from a two-sided t-test of the t-value for the corresponding regression coefficient.