Modelling the distribution of antimicrobial resistance genes across geography, climate, host species and phylogroup in 16,000 E. coli genomes

E. coli is a highly diverse bacterial species that generates a huge global burden of antimicrobial-resistant infections. A wealth of whole genome sequence data is available on public databases for this species, presenting new opportunities to analyse the distribution of antimicrobial resistance (AMR) genes across its genetic and ecological diversity. We extracted and categorised metadata on host species and geographic location and combined this with in silico phylogrouping to describe the characteristics of [~]16,000 assembled E. coli genomes from the NCBI RefSeq database. We estimated AMR carriage using various metrics: counts of overall genes, multidrug- and extensively drug-resistant categories, and selected {beta}-lactamases of current global concern - blaCTX-M and carbapenemase genes. We present estimates of AMR carriage for these metrics by species type (human, agricultural/domestic animal, wild birds and other wild animals), geographic subregion, and across phylogroups. In addition, we describe the distribution of phylogroups within host types and geographic subregions. Our findings show high AMR carriage in commensal-associated phylogroups, agricultural and wild animal hosts and in many subregions. However, we also quantify large biases in sequencing data, the substantial gaps in our knowledge of AMR in many hosts, regions and environmental settings, and the need for systematic sampling to gain a more accurate picture.