Machine learning classification can reduce false positives in structure-based virtual screening.

With the recent explosion in the size of libraries available for screening, virtual screening is positioned to assume a more prominent role in early drug discovery's search for active chemical matter. In typical virtual screens, however, only about 12% of the top-scoring compounds actually show activity when tested in biochemical assays. We argue that most scoring functions used for this task have been developed with insufficient thoughtfulness into the datasets on which they are trained and tested, leading to overly simplistic models and/or overtraining. These problems are compounded in the literature because studies reporting new scoring methods have not validated their models prospectively within the same study. Here, we report a strategy for building a training dataset (D-COID) that aims to generate highly compelling decoy complexes that are individually matched to available active complexes. Using this dataset, we train a general-purpose classifier for virtual screening (vScreenML) that is built on the XGBoost framework. In retrospective benchmarks, our classifier shows outstanding performance relative to other scoring functions. In a prospective context, nearly all candidate inhibitors from a screen against acetylcholinesterase show detectable activity; beyond this, 10 of 23 compounds have IC50 better than 50 µM. Without any medicinal chemistry optimization, the most potent hit has IC50 280 nM, corresponding to Ki of 173 nM. These results support using the D-COID strategy for training classifiers in other computational biology tasks, and for vScreenML in virtual screening campaigns against other protein targets. Both D-COID and vScreenML are freely distributed to facilitate such efforts.

Cellular Growth Arrest and Persistence from Enzyme Saturation.

Metabolic efficiency depends on the balance between supply and demand of metabolites, which is sensitive to environmental and physiological fluctuations, or noise, causing shortages or surpluses in the metabolic pipeline. How cells can reliably optimize biomass production in the presence of metabolic fluctuations is a fundamental question that has not been fully answered. Here we use mathematical models to predict that enzyme saturation creates distinct regimes of cellular growth, including a phase of growth arrest resulting from toxicity of the metabolic process. Noise can drive entry of single cells into growth arrest while a fast-growing majority sustains the population. We confirmed these predictions by measuring the growth dynamics of Escherichia coli utilizing lactose as a sole carbon source. The predicted heterogeneous growth emerged at high lactose concentrations, and was associated with cell death and production of antibiotic-tolerant persister cells. These results suggest how metabolic networks may balance costs and benefits, with important implications for drug tolerance.