Network Analysis of Correlated Mutations in Influenza

Influenza A Virus (IAV) is remarkably adept at surviving in human populations. IAV thrives even among populations with wide spread access to vaccines and anti-viral drugs, and continues to be a major cause of morbidity and mortality. Correlated mutations are an important factor in IAV's evolution and are critical for host adaptation and pathogenicity. Large sets of publicly available sequences of IAV combined with its rapid and complex evolutionary dynamics present interesting opportunities and unique challenges to analyze correlated mutations in influenza proteomes. In this work, we performed a comprehensive analysis of correlated mutations in IAV using a network theory approach where residues in each protein act as nodes in the graph and edges in the graph are created based on inter-residue correlated mutations. Our approach used 'maximal information coefficient' (MIC) to compute correlations between residues and the edges connect nodes if their MIC exceeds a threshold. We created a modular and robust pipeline and applied it to multiple datasets of H1N1, H3N2, H5 and H7N9 subtypes. We studied structural dynamics of IAV sub-systems based on topological properties of their networks resulting in several important conclusions. The main finding is that correlated mutation networks in IAV are sub-type and host specific and the differences for various subtypes and hosts are significant. We identified nodes with highest degree along with edges and triplets with strongest weight for each network. To contextualize our results, we performed entropy analysis to gain a global view of sequence variation and computed solvent accessibility profiles to identify statistical differences in correlation profiles between surface and buried residues. To understand the extent of co-variation between the 10 proteins in IAV sequences, we created visualizations of protein correlation graphs where the proteins acts as nodes and the strength of connections between the nodes depends on the number of correlated mutations between residues of connected proteins. A web application and visualization tools to explore the results and search for correlated mutations were developed.