Detecting signatures of episodic positive selection based on observed amino acids in hemagglutinin of H3N2 human influenza A virus

Abstract

In the parsimony method for detecting natural selection at amino acid sites of proteins, the numbers of synonymous ($d_S$) and nonsynonymous ($d_N$) substitutions that have accumulated over the evolutionary history of observed sequences were computed assuming that any amino acid was compatible at each site. Positive selection was inferred to have operated recurrently when the null hypothesis of no selection was rejected with $d_S<d_N$. Here an attempt to detect episodic positive selection within the framework of parsimony method was demonstrated. The $d_S$ and $d_N$ values were computed assuming that only the observed amino acids were compatible at each site. Positive selection was inferred to have operated episodically when the null hypothesis of the same fitness effects among observed amino acids was rejected with $d_S>d_N$. In the analysis of 18,444 sequences for hemagglutinin of H3N2 human influenza A virus, recurrent and episodic positive selections were inferred mainly at the sites related to antigenicity. Episodic positive selection was detected particularly at the sites under epistasis. Although it may be necessary to eliminate slightly deleterious amino acids from the population genetic data, the analysis based on observed amino acids may be useful for screening the sites with signatures of episodic positive selection.