Global Migration and Transmission Dynamics of Influenza H3N2

Abstract

Phylogeographic methods have been used extensively to characterize the global migration patterns of in- fluenza. Those methods however make strong assumptions about independence of tree topology and migra- tion (Lemey et al., 2009) or assume populations to be constant over time (Vaughan et al., 2014). We imple- mented an approximate structured coalescent method (Volz et al., 2009; Volz, 2012) in BEAST 2 (Bouckaert et al., 2014) that relaxes these assumptions by allowing for seasonally changing transmission rates. We show first on simulated data that our approximate structured coalescent infers asymmetries in migration rates better compared to (Lemey et al., 2009) on a variety of different parameter ranges. We then show on globally sampled influenza H3N2 data that the choice of model can have great influence on the inference of the location of the trunk, i.e. the backbone, of the phylogeny of H3N2. Consistent with previous find- ings (Bedford et al., 2010; Le et al., 2013; Bedford et al., 2015), we find that China, India and New York are likely sources of H3N2. Using our newly implemented method, we find that South East Asia is likely to play a smaller role in the global migration network of H3N2 compared to previous work e.g. (Lemey et al., 2014).