Integrated deep learned transcriptomic and structure-based predictor of clinical trials outcomes

摘要: Despite many recent advances in systems biology and a marked increase the availability of high-throughput biological data, productivity research development pharmaceutical industry is on decline. This primarily due to clinical trial failure rates reaching up 95% oncology other disease areas. We have developed comprehensive analytical computational pipeline utilizing deep learning techniques novel tools predict outcomes phase I/II trials. The predicts side effects drug using neural networks estimates drug-induced pathway activation. It then uses predicted effect probabilities activation scores as an input train classifier which outcomes. was trained 577 transcriptomic datasets has achieved cross-validated accuracy 0.83. When compared direct gene-based classifier, our multi-stage approach dramatically improves predictions. applied set compounds currently present pipelines several major companies highlight potential risks their portfolios estimate fraction trials that were likely fail I II.