Deep-learning single-cell profiling uncovers interbacterial and drug-driven reshaping of pathogen virulence

Abstract Bacterial invasion and genotoxicity are key contributors to infectious disease progression and carcinogenesis. Yet current methods do not quantify these processes systematically and at single-cell resolution, limiting our ability to explore how interplay of virulence factors and drug effects shape pathogenicity. Here, we present MALVINA, a high-throughput deep-learning workflow integrating host-pathogen co-culture and high-content imaging to jointly measure bacterial invasion and host DNA damage. Applied to Escherichia coli isolates and human colorectal epithelial cells, we demonstrate MALVINA’s versatility by deciphering (1) strains with elevated invasive or genotoxic potential, (2) interbacterial modulation of virulence, (3) pro-virulent drug effects, (4) virulence-targeting drugs, and (5) bacterial potentiation of drug toxicity. MALVINA revealed strain-specific virulence profiles and colibactin-dependent suppression of competitors’ invasion coupled with enhanced genotoxicity. It also enabled identification of drug-induced modulation of bacterial invasion and genotoxicity, offering a new perspective for virulence-aware drug profiling. Overall, MALVINA offers an adaptable framework for dissecting pathogen-host-drug interactions.