Defining permissive chemical space in Klebsiella pneumoniae

Our current failure to understand the chemical properties responsible for compound permeation, retention, and metabolism in bacteria is compromising efforts to develop new antibiotics particularly against novel targets. Our proposal will leverage our existing methods developed to experimentally map (and subsequently predict) compound retention and metabolism in Mycobacteria to: (i) define the chemical constrains for antibiotics in Klebsiella pneumoniae (Kp) by screening a 10,000-compound library for retention and metabolism, refine boundary conditions by chemical elaboration, and then use this dataset to build ML predictive models for unseen molecules; (ii) understand the molecular basis for chemical retention and metabolism by testing an informative compound collection against existing arrayed deletion mutants and newly generated CRISPRi knockdown strains targeting routes of entry (porins, transporters), metabolic enzymes, and efflux pumps; (iii) explore variation of chemical retention and metabolism across an existing library of 500 genome-sequenced clinical isolates of Kp spanning phylogenetic diversity, to define invariant permissive chemical space and, through GWAS, examine the genetic basis for variation; and (iv) create open access, web- based ML tools to predict retention/metabolism of compounds (and thereby guide phenotypic hit-to-lead optimisation, in silico screening, and fragment elaboration efforts) and create permeability-aware generative AI methods for antibiotic discovery.