Predicting Drug Side Effects with Advanced Deep Learning: The DeepSide Approach

Abstract

Unexpected adverse effects that cause clinical trial drug failures endanger participant health and cause large financial losses. Potentially, side effect prediction algorithms could guide the development of new drugs. The LINCS L1000 dataset provides a wealth of cell line gene expression data disrupted by different drugs and establishes a foundation of knowledge for context-specific factors. The state-of-the art approach uses just the high-quality trials in LINCS L1000, discarding a large proportion of the tests to employ context specific information.