Sarand: exploring antimicrobial resistance gene neighbourhoods in complex metagenomic assembly graphs.

Antimicrobial resistance (AMR) is a major global challenge to human and animal health. The genomic element (e.g. chromosome, plasmid, and genomic islands) and neighbouring genes associated with an AMR gene play a major role in its function, regulation, evolution, and propensity to undergo lateral gene transfer. Therefore, characterizing these genomic contexts is vital for effective AMR surveillance, risk assessment, and stewardship. Metagenomic sequencing is widely used to identify AMR genes in microbial communities but fragmentary short-read data do not directly provide this critical contextual information. Assembly of these reads provides some contextual information but fails to recover many mobile genetic elements. Here, we introduce Sarand, a method retaining some of the sensitivity of read-based methods while providing the genomic context of assembly by extracting AMR genes and their associated context directly from metagenomic assembly graphs. Sarand uses BLAST-based homology searches with coverage statistics to identify and visualize AMR gene contexts while filtering false chimeric contexts. Using both real and simulated metagenomic data, we show that Sarand outperforms metagenomic assembly and other recently developed graph-based tools in terms of precision and sensitivity for this problem. Sarand enables effective extraction of metagenomic AMR gene contexts to better characterize AMR evolutionary dynamics within complex microbial communities.