Fig. 3: Lagging-strand genes are more frequently under positive selection.
From: Reply to: Testing the adaptive hypothesis of lagging-strand encoding in bacterial genomes
Likelihood ratio tests (LRTs) were used to compare site models M1a (neutral model) versus M2a (positive selection model) as a test for positive selection among genes with a dN/dS ratio >1. The frequency of genes better fitting the positive selection model at >95% significance are plotted (# genes under positive selection/total). Positive selection likely acts more frequently on lagging-strand genes in M. tuberculosis (Fisher’s exact test p = 0.049). An equivalent analysis is shown for B. subtilis (middle graph, Fisher’s exact test p = 0.12 for leading versus lagging-strand frequencies). Multi-species analysis: across species, genes under positive selection are more frequent in lagging-strand genes, confirming the results of our original analysis (right graph, Fisher’s exact test p = 0.005).
