Genetic programming in data mining for drug discovery

Genetic programming (GP) is used to extract from rat oral bioavailability (OB) measurements simple, interpretable and predictive QSAR models which both generalise to rats and to marketed drugs in humans. Receiver Operating Characteristics (ROC) curves for the binary classier produced by machine learning show no statistical dierence between rats (albeit without known clearance dierences) and man. Thus evolutionary computing oers the prospect of in silico ADME screening, e.g. for \virtual" chemicals, for pharmaceutical drug discovery