Secondary pharmacology profiling using quantitative structure-activity relationship (QSAR) models offer an efficient way to provide insight into biological properties during compound optimisation and prioritisation. The performance of these models is highly dependent on the quality and the quantity of the data available to train them, particularly when investigating new areas of chemical space.
A wealth of knowledge is locked in high-quality proprietary data silos and allowing QSAR models to access this knowledge would lead to unprecedented performances and decision support. Federated learning can overcome the confidentiality through using a privacy-preserving approach to extract knowledge from proprietary data and facilitate pre-competitive collaboration.