Adding stochastic negative examples into machine learning improves molecular bioactivity prediction

Multitask deep neural networks learn to predict ligand-target binding by example, yet public pharmacological datasets are sparse, imbalanced, and approximate. We constructed two hold-out benchmarks to approximate temporal and drug-screening test scenarios whose characteristics differ from a random split of conventional training datasets. We developed a pharmacological dataset augmentation procedure, Stochastic Negative Addition (SNA), that randomly assigns untested molecule-target pairs as transient negative examples during training. Under the SNA procedure, ligand drug-screening benchmark performance increases from R2 = 0.1926 {+/-} 0.0186 to 0.4269{+/-}0.0272 (121.7%). This gain was accompanied by a modest decrease in the temporal benchmark (13.42%). SNA increases in drug-screening performance were consistent for classification and regression tasks and outperformed scrambled controls. Our results highlight where data and feature uncertainty may be problematic, but also show how leveraging uncertainty into training improves predictions of drug-target relationships.