Host-associated and free-living phage communities differ profoundly in phylogenetic composition

Phylogenetic profiling has been widely used for comparing bacterial communities, but has so far been impossible to apply to viruses because of the lack of a single marker gene analogous to 16S rRNA. Here we developed a reference tree approach for matching viral sequences and applied it to the largest viral datasets available. The resulting technique, Shotgun UniFrac, was used to compare host-associated and non-host-associated phage communities (130 total metagenomes), and revealed a profound split similar to that found with bacterial communities. This new informatics approach complements analysis of bacterial communities and promises to provide new insights into viral community dynamics, such as top-down versus bottom-up control of bacterial communities by viruses in a range of systems.