Search for a command to run...
Tackling the Gram-Negative Permeability Barrier Through Collaborative Data-Driven Prediction Gram-negative membrane permeability is a major bottleneck in antibiotic discovery. We team up to combine models and data for improved permeability prediction and compound prioritisation. Limited membrane permeability is a key reason why many compounds fail to show activity in Gram-negative pathogens. Large screening libraries are available, but identifying compounds that can effectively enter Gram-negative cells remains challenging. Experimental permeability measurements are scarce, heterogeneous, and costly, limiting their use at scale. Within the AMR Accelerator, the COMBINE project coordinates an established Scientific Interest Group in Machine Learning, bringing together academic and industry partners to develop predictive models for Gram-negative permeability. Building on existing work by partners (Gadiya et al. 2025; Le Goff et al. 2025; Scalia et al. 2025), we combine complementary datasets and tools to enable the in silico prioritisation of compounds from large screening libraries. By pooling resources across the AMR Accelerator community, we aim to reduce time and cost in early-stage antibiotic discovery and improve our understanding of the determinants of compound uptake in Gram-negative bacteria. Call for data The AMR Accelerator is looking for data. We need YOUR help! Following our successful Antibacterial property prediction ML model we now seek experimental datasets that inform on compound permeability in Gram-negative bacteria. This includes direct measurements such as uptake or accumulation assays, as well as relevant proxy readouts (e.g. MIC-based data) where direct permeability measurements are not available. We also welcome expertise, tools, and methodologies related to permeability assessment, data analysis, or predictive modeling. Previous work using public data sets to build Antibacterial property prediction ML models. See DOI: 10.1021/acs.jcim.4c02347 Connect with COMBINE Contact Leonie von Berlin (Fraunhofer ITMP, leonie.von.berlin@itmp.fraunhofer.de) or scan the QR code for more information!