The strategy is to simultaneously optimize the quantified specificity of the ‘‘native’’ protein-ligand complex discriminating against ‘‘non-native’’ binding modes and the affinity prediction. The benchmark testing of SPA shows the best performance against 16 other popular scoring functions in industry and academia on both prediction of binding affinity and ‘‘native’’ binding pose. For the target COX-2 of nonsteroidal anti-inflammatory drugs, SPA successfully discriminates the drugs from the diversity set, and the selective drugs from non-selective drugs. The remarkable performance demonstrates that SPA has significant potential applications in identifying lead compounds for drug discovery.
This article was published online in Scientific Reports and is free to access.