Learning Natural Selection from the Site Frequency Spectrum

Supporting Information

Supporting Information

Supporting Information - File S1, Figures S1-S10, and Tables S1-S3 (PDF, 3 MB)
File S1 - Supporting Information (PDF, 84 KB)
Figure S1 - Power (0.05 FPR) of the cross-population SVM test compared to other cross-population tests of neutrality. (PDF, 307 KB)
Figure S2 - Pairwise cosine distance between 200 SVMs trained on cross-population data (matrices of the XP-SFS scaled to 8 x 8 frequency bins, and vectorized). (PDF, 2 MB)
Figure S3 - Feature weights learned from the XP-SFS on data corresponding to the two observed regimes of selection: (A) near-fixation, and (B) post-fixation. (PDF, 396 KB)
Figure S4 - Power (0.05 FPR) of neutrality tests based on supervised learning. (PDF, 288 KB)
Figure S5 - XP-SFselect values on Drosophila chromosome 2L (PDF, 376 KB)
Figure S6 - XP-SFselect values on Drosophila chromosome 2R (PDF, 376 KB)
Figure S7 - XP-SFselect values on Drosophila chromosome 3L (PDF, 389 KB)
Figure S8 - XP-SFselect values on Drosophila chromosome 3R (PDF, 389 KB)
Figure S9 - Power of SFselect using SVM and logistic regression, at different times and selection pressures (PDF, 235 KB)
Figure S10 - Power of XP-SFselect using SVM and logistic regression, at different times and selection pressures (PDF, 235 KB)
Table S1 - List of significant regions under XP-SFselect for the fly hypoxia experiments described in Zhou et al. (2011). (PDF, 171 KB)
Table S2 - The top 40 non-overlapping regions identified genome-wide by XP-SFselect (PDF, 192 KB)
Table S3 - Potentially damaging SNPs found in regions with strong evidence of non-neutral evolution (PDF, 222 KB)