Fig. 3

Interferon-stimulated genes are enriched for sequence signatures of recurrent positive natural selection. a Graphical illustrations of the M8 and M8a nested codon models in PAML (Yang 2007). M8a is a null model where all codons in the multiple sequence alignment must be placed into one of 11 bins of specific dN/dS values. Ten of these bins are distributed along a beta distribution of dN/dS values bounded between 0 and 1. The 11th bin is defined to have a dN/dS value = 1. M8 is identical, except that the 11th bin can have a dN/dS value greater than one [43]. The double-sided arrow indicates that the dN/dS value of this bin is optimized in the fitting of the data to the model. b 331 gene alignments were fit to both the M8 and the M8a models. A likelihood ratio test of the two nested models was conducted, and the final column indicates the number of genes in each category for which the null model M8a could be rejected in favor of the model of positive selection (p < 0.05). We did a Benjamini–Hochberg correction at 10% FDR to control for false positives. In the pie charts, the proportion of genes in each category that are under positive selection (red) is shown, from the table above. Using a two-tailed Fisher's exact test and Benjamini–Hochberg correction at 10% FDR, the difference in the number of genes rejecting the neutral model (M8a) between random genes and interferon-stimulated genes was significant