AI/ML and Molecular Dynamics in the Discovery of Efflux Pump Inhibitors

Efflux pump research spans antimicrobial resistance and cancer drug discovery, as these transporters can reduce intracellular drug concentrations and contribute to treatment failure. Transporters from the MFS, MATE, SMR and ABC superfamilies recognize diverse compounds, while their changing conformations complicate inhibitor discovery. AcrAB-TolC, MexAB-OprM, NorA and P-glycoprotein illustrate the range of systems under investigation. Structural studies have expanded knowledge of these transporters, but static structures alone cannot capture transport cycles involving substrate recruitment, occlusion, translocation and release. No efflux pump inhibitor has reached clinical use. Molecular dynamics and AI/ML methods can help investigate transporter dynamics, binding energetics and inhibitor design, but findings may be limited by sparse experimental data, sampling challenges and model uncertainty. Further work is needed to connect computational predictions to experimental evidence and to establish clear, reproducible standards for evaluating them.This Research Topic aims to advance in silico drug discovery for efflux pump inhibitors by integrating molecular dynamics, computational chemistry, structural bioinformatics and AI/ML with experimental data. It seeks to improve understanding of transporter conformational landscapes, substrate and inhibitor interactions, translocation energetics and allosteric communication, and to assess methods for predicting or designing potential inhibitors. New laboratory experiments are not required for every submission, but studies must be grounded in experimental data - for example, experimentally determined structures or published binding, activity, MIC, accumulation or mutagenesis results. For ML predictive models, new independent experimental validation is not mandatory if the models are rigorously evaluated using experimental data, appropriate held-out or external testing, relevant baselines and controls, and consideration of data leakage and uncertainty. For generative and other de novo design studies, chemical synthesis and biological testing are not mandatory; proposed compounds should be presented as predicted candidates and assessed in relation to existing experimental evidence. Claims of demonstrated inhibitory activity, or hypotheses about transporter function or inhibition in biological systems, must be supported by appropriate experimental evidence. Negative, inconclusive and cautionary findings are welcome.To gather further insights into computational methods and AI for drug discovery targeting bacterial, fungal, parasitic and mammalian efflux transporters, this Research Topic focuses on studies with a strong in silico component and a clear connection to experimental data; purely theoretical or computational studies without such a connection, and primarily experimental studies without a substantial computational component, are outside its scope. New experiments are not required in every submission, but experimental evidence must inform the study’s data, evaluation or interpretation. We welcome articles addressing, but not limited to, the following themes:&#8226 Machine-learning methods, including graph neural networks and transformer models, for substrate or inhibitor prediction&#8226 Generative chemistry, de novo design, structure-generation, and structure-based or ligand-based design of potential efflux pump inhibitors&#8226 Virtual screening, small-molecule docking and computer-assisted compound selection for efflux substrates and inhibitors&#8226 ADME/Tox modelling and prediction for candidate compounds&#8226 Computational evaluation of predictive and generative models, including experimental-data benchmarks, external validation, data leakage, uncertainty and reproducibility&#8226 Molecular dynamics simulations in membrane environments, including assessment of sampling, convergence and force-field limitations&#8226 Coarse-grained modelling, correlation analysis and allosteric-network characterization&#8226 Protonation-coupled conformational transitions and transport-cycle modelling&#8226 Enhanced-sampling approaches, including metadynamics, umbrella sampling, replica exchange and adaptive sampling&#8226 Markov state models and other approaches to resolving conformational landscapes and transition pathways&#8226 Alchemical and pathway-based free-energy calculations of binding and translocation&#8226 Integrative computational–experimental studies using approaches such as MIC assays, intracellular accumulation measurements, mutagenesis or cryo-electron microscopy, with experimental evidence appropriate to the claims made Topic editors: Nandan Kumar, Vassiliy Bavro, Sita Sirisha Madugula Source: https://www.frontiersin.org/research-topics/85192/aiml-and-molecular-dynamics-in-the-discovery-of-efflux-pump-inhibitors