High throughput genetic sequence analysis

Abstract

We present an application paradigm in which an unsupervised machine learning approach is applied to high dimensional influenza sequence datasets: (1) human A/H3N2, (2) avian H5, and (3) North American swine influenza H3N2 virus. Interesting visual patterns observed in the A/H3N2 influenza virus led us to hypothesize that vaccination could be one of the driving forces in the evolution of the human A/H3N2 influenza virus. We provide simulation study and statistical results to support our finding that the influenza virus evolves differently in a protected environment than it evolves in the wild. In the swine H3N2 case, our result suggests that the diversification of North American swine influenza virus can be attributed to the mutations at two positively selected sites on the hemaggluttinin protein.