|
Abstract:
|
With the advent of cheap sequencing, species variation is increasingly visible in space and over time. In particular, our interest focuses on virus species observed to differentiate into populations that co-exist or appear sequentially in epidemics or within hosts. Phylogenetic methods can illuminate such population structure in sequence samples, but typically resolve much finer structure than may be justified. Population structure inference methods also apply, but they make very strong independence assumptions. We describe an EM algorithm for inferring phyloclusters, that is, evolutionarily related clusters of sequences. We provide rigorous methods for initialization, estimation of the number of clusters, and hypothesis testing. We apply the method to several viral data sets, comparing our findings to previous results based on phylogenetic and other analyses.
|
