Computational Genomics for Microbial Traits

About the speaker:

Dr Janani Ravi is an Assistant Professor at CU Anschutz, Department of Biomedical Informatics, Centre for Health Artificial Intelligence. With a PhD in Computational Biology from Virginia Tech and postdoctoral research at Rutgers, she joined CU in Fall 2022 from Michigan State University. Her group develops computational approaches integrating large-scale public datasets to enable mechanistic understanding of microbial genotypes, phenotypes, and diseases. Her first research aim is to understand how microbes adapt to diverse environments and how their genetic makeup relates to observable traits, using protein sequence-structure-function relationships, comparative genomics, and machine learning to bridge genotype and phenotype, with a focus on antimicrobial resistance, host specificity, microbial pathogenesis, and plant-soil-microbe adaptation. Her second research aim seeks to identify the molecular mechanisms underlying host responses to infection and to discover therapeutics targeting these responses.

Abstract:
Microbes span a broad spectrum, ranging from free-living organisms to obligate pathogens, yet the precise genomic mechanisms underlying niche adaptation and pathogenesis remain poorly characterised. This work presents general-purpose computational approaches that integrate large-scale heterogeneous public datasets to bridge the microbial genotype–phenotype gap. By leveraging protein sequence–structure–function relationships, comparative genomics, and machine learning (ML) across multiple molecular scales, including genes, k-mers, and proteins, this framework demonstrates how microbial adaptation and phenotypic outcomes can be systematically studied. Interpretable ML models were developed to predict antimicrobial resistance (AMR) across 60 drugs, achieving high performance (median normalised MCC = 0.89), while recovering known resistance mechanisms and identifying novel candidate determinants potentially driven by horizontal gene transfer. Extending this framework, comparative transcriptomics and disease–drug signature analyses were employed to delineate host responses to infection and to accelerate drug repurposing for tuberculosis and other infectious diseases. The proposed methods are microbe-, host-, and disease-agnostic, and are accompanied by open-source software and interactive web tools to facilitate broad community reuse.