부산대학교 도서관

Various computational and statistical approaches have been proposed to uncover the mutational patterns of rapidly evolving influenza viral genes. A problem that draws much attention is to identify pairs of sites that potentially co-mutate to contribute to the overall fitness of the virus. Unlike previous methods that extract the mutations from sequence alignments, here we endeavor a novel method that relies on identifying mutations in the phylogenetic trees that are reconstructed using resampled sequence data. Since the method takes into account the evolutionary structure presents in the sequence data, spurious mutations obtained by comparing sequences from different clades could be removed. Furthermore, this approach does not only allow us to capture site-pairs that potentially co-mutate, but also provides an opportunity to extract the direction of their relationships. By applying network analyses to the set of site-pairs, we could further identify and rank the sites that are likely to be influential or under influence from changes on other sites. We applied the method to the hemagglutinin of influenza H3N2, and interestingly, we successfully recovered mutational sites that are important for cluster antigenic transition of the virus in the top list of our findings. Moreover, we detected a directional relationship that would be interesting for experimental investigation.