Modeling ecological drivers in marine viral communities using comparative metagenomics and network analyses

Microorganisms and their viruses are increasingly recognized as drivers of myriad ecosystem processes. However, our knowledge of their roles is limited by the inability of culture-dependent and culture-independent (e.g., metagenomics) methods to be fully implemented at scales relevant to the diversity found in nature. Here we combine advances in bioinformatics (shared k-mer analyses) and social networking (regression modeling) to develop an annotation- and assembly-free visualization and analytical strategy for comparative metagenomics that uses all the data in a unified statistical framework. Application to 32 Pacific Ocean viromes, the first large-scale quantitative viral metagenomic dataset, tested existing and generated further hypotheses about ecological drivers of viral community structure. Highly computationally scalable, this new approach enables diverse sequence-based large-scale comparative studies.