Global pathogenomic analysis identifies known and novel genetic antimicrobial resistance determinants in twelve species

Abstract

Surveillance programs for managing antimicrobial resistance (AMR) have yielded thousands of genomes suited for data-driven mechanism discovery. We present a workflow integrating pangenomics, gene annotation, and machine learning to identify AMR genes at scale. Applied to 12 species, 27,155 genomes, and 69 drugs, we 1) found AMR gene transfer mostly confined within related species, with 925 genes in multiple species but just eight in multiple phylogenetic classes, 2) demonstrated that discovery-oriented support vector machines outperform contemporary methods at recovering known AMR genes, recovering 263 genes compared to 145 by Pyseer, and 3) identified 142 novel AMR gene candidates. Validation of two candidates in E. coli BW25113 revealed cases of conditional resistance: ΔcycA conferred ciprofloxacin resistance in minimal media with D-serine, and frdD V111D conferred ampicillin resistance in the presence of ampC by modifying the overlapping promoter. We expect this approach to be adaptable to other species and phenotypes.

Alleleomes characterize the survivors of 3.5 billion years of bacterial evolution

“…Furthermore, since we used a direct-sequence-match (to previously identi ed AMR genes) criteria to determine AMR genes in our dataset, it is likely that we underestimate a large fraction of AMR genes that share high sequence similarity with (but are not identical to) our previously identi ed AMR allele sequences. We expect this effect is more pronounced in species that are distantly related to the 12 species for which AMR genes were previously identi ed 20 .…”

“…The alleleomes used in this study were adapted from our previous work 20 , constructed as follows. First, genomes from the Web of Life collection 40 were ltered on the following criteria: 1) GTDB species classi cation (release 202) is available 41 , 2) CheckM 42 contamination < 10% and completeness > 80%, 3) number of contigs is within three times the median number of contigs for all assemblies for the genome's species, and 4) total assembly length is within three standard deviations from the mean of all assemblies for the genome's species.…”

“…Thus, for any gene(s) in any one (or across multiple) species of bacteria, the alleleome can measure the relative strength and mode of the selection pressures acting upon any gene or subset of genes. As an example, AMR-conferring genes 20 are shown in Fig. S6a-b.…”

“…We previously identi ed 69,607 gene-variants of 7,710 AMR-conferring genes 20 . Across all alleles of all strains, we nd 43,042 direct sequence matches to this set of gene-variants, representing unique 3,757 AMR genes across 141 species.…”