Fig. 1: Overview of Argo.

a Screening of ARG reads. Given input long reads, those that carry at least one ARG are extracted using DIAMOND’s frameshift-aware DNA-to-protein alignment with an adaptive identity cutoff. b Collection of host information. ARG-containing reads are mapped to an ARG-tailored reference taxonomy database (GTDB) to obtain candidate species sets and alignment scores using minimap2’s base-level alignment. c Assignment of taxonomic labels. ARG-containing reads are overlapped with each other to form an overlap graph using minimap2’s approximate mapping (without base-level details). The graph (each vertex represents a read, and each edge the identity between reads) is segmented into components (read clusters) using the MCL algorithm. Taxonomic labels of reads within each cluster are determined by solving a weighted set cover problem with a greedy approximation. The final output of Argo is a tab-delimited table listing ARG abundances by species. a–c Semi-transparent gray dots stand for ARGs. Colors indicate different species.