Chapter 15: Data-driven Prediction of Organic Reaction Outcomes

Anticipating the outcome of a chemical reaction is a task that chemists routinely perform when designing syntheses and synthetic routes. Computational approaches have evolved over several decades from expert systems—where rules and patterns of reactivity are painstakingly encoded by hand—to data-driven systems—where similar patterns are instead inferred from data. There is now a host of algorithms for reaction prediction with various problem formulations: with or without reaction templates, operating at the mechanistic or global reaction level, and using molecules represented as fingerprints, graphs, or even SMILES strings. This chapter highlights recent progress in the field of data-driven reaction prediction with an emphasis on emerging machine learning techniques.