Capturing variation in metagenomic assembly graphs with MetaCortex
The assembly of contiguous sequence from metagenomic samples presents a particular challenge, due to the presence of multiple species, often closely related, at varying levels of abundance. Capturing diversity within species, for example viral haplotypes, or bacterial strain-level diversity, is even more challenging. We present MetaCortex, a metagenome assembler based on data structures from the Cortex de novo assembler. MetaCortex captures intra-species diversity by searching for signatures of local variation along assembled sequences in the underlying assembly graph and outputting these sequences in sequence graph format. MetaCortex also implements a novel assembly algorithm for representing intra-species diversity in standard linear format. We show that MetaCortex produces accurate assemblies with higher genome coverage and contiguity than other popular metagenomic assemblers on mock viral communities with high levels of strain level diversity, and on simulated communities containing simulated strains. We also show that accuracy can be increased further by using the sequence graph produced by MetaCortex to create highly accurate single contig sequences.