NLM DIR Seminar Schedule

Abstract:

Microbes exist across a spectrum, from free-living environmental organisms to obligate intracellular pathogens, yet the precise genomic mechanisms underlying niche adaptation and pathogenesis remain poorly characterized in most species. We develop general-purpose computational approaches that integrate large-scale heterogeneous public datasets to bridge the microbial genotype-phenotype gap. Using protein sequence-structure-function relationships, comparative genomics, and machine learning (ML) across multiple molecular scales (genes, k-mers, proteins, domains), we study how microbes adapt to distinct niches and how their genotypes determine phenotypic outcomes. Focusing on clinically critical pathogens, including the ESKAPE pathogens, our interpretable ML models predict antimicrobial resistance (AMR) across ~60 drugs and 6 taxa with high performance (median normalized MCC=0.89), recover known resistance mechanisms, and discover novel candidates driven by horizontal gene transfer. Extending this framework, we predict host specificity in S. aureus, implicating the prophage ϕSa3 immune evasion cluster as a key determinant of human host preference. Complementing the microbial side, we also apply comparative transcriptomics and disease-drug signatures to delineate host responses to infection and accelerate drug repurposing (host-directed therapeutics) for tuberculosis and beyond. Our methods are microbe-, host-, and disease-agnostic; we release open data, software, and interactive web tools to enable broad community use and reuse.