Author: Kristopher  Kieft,  Graduate Student (University of Wisconsin-Madison)

Description:
Viruses are central to microbial communities and the cycling of nutrients in all environments. As many as 30-40% of all bacteria are infected by viruses at any given time and undergoing viral-directed allocation of nutrients, such as carbon and nitrogen processing. The ability to generate large metagenomic assemblies of mixed microbial and viral sequences provides the opportunity to tease apart complex microbiome dynamics, but these analyses are limited by the tools available for analyses of viral genomes. Here we present VIBRANT, the first method to utilize a hybrid machine learning and protein similarity approach that is not reliant on sequence features for automated recovery and annotation of viruses, determination of genome quality, and characterization of viral community function from metagenomic assemblies or genome sequences. VIBRANT uses neural networks of protein signatures and a novel v-score metric to maximize identification of lytic viral genomes and integrated proviruses, including highly diverse viruses. VIBRANT highlights viral auxiliary metabolic genes, thereby serving as a user-friendly platform for evaluating viral metabolic functions. VIBRANT was trained and validated on reference virus datasets as well as microbiome and virome data from a diversity of environments, such as marine and freshwater. The ability to incorporate viruses into the advent of big data generation, such as long-term timeseries studies or whole community sequencing, will greatly advance our understanding of microbiomes, host-microbe interactions and ecosystem dynamics.

Category: Scientific Program Abstract > Special Session > CS29 Viruses