Large-Scale Comparative Genomics and Structure-Function Analysis Enables Characterization of Known and Novel Genetic Determinants of Antimicrobial Resistance in Bacterial Pathogens
Antibiotics are crucial for preventing infection-induced complications, but their widespread overuse has spurred the evolution of antimicrobial resistance (AMR) mechanisms in pathogens. Data-driven biosurveillance approaches utilizing whole genome sequencing data and computational approaches have the potential to improve the detection and characterization of known and emerging AMR profiles, especially in high-priority ESKAPE, enteric, and sexually-transmitted pathogens. In this study, we performed a large-scale analysis of over 70,000 genomes representing 39 pathogen-antibiotic combinations to identify resistance determinants statistically enriched in antibiotic resistant strains. Using a kmer-based GWAS approach, over 7,000 unique sequences were identified among all resistant genomes. Of these, 1,925 sequences were homologous to known AMR genes, while over 5,000 sequences lacked homology, suggesting novel AMR-associated genes. In addition to identifying the predominant AMR genes for specific pathogen-antibiotic combinations, our findings suggest that horizontal gene transfer mechanisms may influence AMR gene profiles between phylogenetically similar pathogens and antibiotic classes. Likewise, significant associations in co-harbored, multi-drug resistance mechanisms were identified in select pathogens. Protein domains analysis frequently detected efflux/membrane structure and antibiotic-associated metabolism domains in novel AMR-associated proteins, suggesting additional mechanisms potentiate resistance phenotypes. Furthermore, we developed a Random Forest classifier using protein structure, molecular features, and binding affinity profiles to predict protein-antibiotic interactions, identifying several novel proteins that may interact with antibiotics. This study demonstrates the potential of large-scale comparative genomics coupled with AI/ML-based modeling to advance the understanding of AMR threats, thereby enhancing biosurveillance efforts and promoting new strategies to counteract emerging pathogens.