10 April 2025 in Nature Communications
The rise in antimicrobial resistance poses a worldwide threat, reducing the efficacy of common antibiotics. Determining the antimicrobial activity of new chemical compounds through experimental methods remains time-consuming and costly. While compound-centric deep learning models promise to accelerate this search and prioritization process, current strategies require large amounts of custom training data. Here, we introduce a lightweight computational strategy for antimicrobial discovery that builds on MolE (Molecular representation through redundancy reduced Embedding), a self-supervised deep learning framework that leverages unlabeled chemical structures to learn task-independent molecular representations. By combining MolE representation learning with available, experimentally validated compound-bacteria activity data, we design a general predictive model that enables assessing compounds with respect to their antimicrobial potential. Our model correctly identifies recent growth-inhibitory compounds that are structurally distinct from current antibiotics. Using this approach, we discover de novo, and experimentally confirm, three human-targeted drugs as growth inhibitors of Staphylococcus aureus. This framework offers a viable, cost-effective strategy to accelerate antibiotic discovery.
1 October 2026 in bioRxiv
Human-targeted drugs are increasingly recognized to affect bacterial physiology, yet their impact beyond growth inhibition, and the underlying molecular mechanisms, remain poorly understood. Here, we systematically mapped how chemically diverse host-associated compounds influence virulence and intrinsic antibiotic resistance in Salmonella enterica serovar Typhimurium. Using a high-throughput reporter approach, we quantified transcriptional response of 26 key stress promoters in Salmonella against 2,415 compounds, including FDA-approved drugs, food-derived molecules, and human metabolites. We found that non-antibiotic compounds induce transcriptional responses distinct from those elicited by antibiotics, uncovering a regulatory shift from general stress responses to compound-specific regulation driven by ligand-responsive MarR- and TetR-family regulators. In particular, more than 200 non-antibiotic drugs modulate virulence or intrinsic resistance pathways through the central regulators SlyA or RamR, respectively. By integrating our large-scale dataset with machine learning, we further defined chemical features associated with ramA transcriptional activation in Salmonella , and predicted ∼200 additional candidate inducers of intrinsic antibiotic resistance in this pathogen. Together, our results uncovered a large and previously unrecognized chemical landscape that reshapes bacterial stress responses through ligand-responsive regulatory mechanisms, with implications for how host-associated compounds influence pathogen virulence and antibiotic resistance. Highlights: Mapping the transcriptional response of key Salmonella stress pathways across a large chemical space shows that antibiotics and non-antibiotic compounds elicit distinct stress responses in Salmonella The activity of the virulence regulator SlyA is modulated by non-antibiotic compounds SlyA-inhibitors increase sensitivity to polymyxin B and decrease Salmonella virulence in cell culture Human-targeted drugs increase intrinsic antibiotic resistance via RamR-RamA in Salmonella Machine learning approach enables large-scale, out-of-sample prediction of human-targeted drugs inducing ramA expression in Salmonella
26 August 2024 in bioRxiv
Transport of small molecules across the bacterial cell envelope is essential to ensure nutrient uptake and protect bacteria from toxic compounds. Due to its decisive role, transport is controlled by complex regulatory networks, knowledge of which – in particular across external cues – remains poorly understood. Here we investigate transcriptional regulation of seven prominent transport genes in Escherichia coli across 94 defined chemical cues, and simultaneously map the contributions of the key regulators MarA, SoxS and Rob to promoter activity. One third of all tested compounds triggered transcriptional changes, the vast majority of which previously unknown. Importantly, we exposed main drivers of transport control in E. coli , e.g. bacteriostatic, but not bactericidal, antibiotics trigger expression of efflux pumps, and that Rob contributes to ∼1/3 of all measured transcriptional changes, thereby emerging as a more prominent transport regulator than previously thought. We showcase the potential of our resource by elucidating the molecular mechanism of antibiotic antagonisms with widely consumed caffeine in E. coli . Altogether, our resource capitalizes on providing a quantitative overview of transport determinants across environments, and brings perspective to long-term prevailing concepts in the field.