A high-throughput screening method for selecting feature SNPs to evaluate breed diversity and infer ancestry

Evaluation of feature SNP selection performance among HITSNP and four other filtration methods. (A) Accuracy of five classifiers based on unfiltered SNP set and different feature SNP sets (about 1000 sites) selected by HITSNP and four other filtration methods. (B) Accuracy of five classifiers based on unfiltered SNP set and different feature SNP sets (about 30,000 sites) selected by HITSNP and four other filtration methods. (C) D_m results of unfiltered SNP data set and feature SNP sets selected by HITSNP and four other filtration methods on about 1000 and about 30,000 SNP sites panels. (D) Accuracy of different feature SNP sets selected by HITSNP and four other filtration methods on diverse gradients of SNP sets. (E) D_m results of different feature SNP sets selected by HITSNP and four other filtration methods on diverse gradients of SNP sets. (F) Jaccard similarity coefficient results of different feature SNP sets selected by HITSNP and four other filtration methods on diverse sizes of SNP sets. (Panels D–F use the same figure legends; “the number of feature SNPs” represents the average number of feature SNPs across five cross-validation folds for the same gradient of SNP sets.)