Discovery of kinase inhibitors by high-throughput docking and scoring based on a transferable linear interaction energy model.

The linear interaction energy method with continuum electrostatics (LIECE) is evaluated in depth on five kinases. The two multiplicative coefficients for the van der Waals energy and electrostatic free energy are shown to be transferable among different kinases. Moreover, good enrichment factors are obtained for a library of 40375 diverse compounds seeded with 73 known inhibitors of CDK2. Therefore, a general two-parameter LIECE model for kinases is derived by combining large data sets of inhibitors of CDK2, Lck, and p38. This two-parameter model is cross-validated on two kinases not used for fitting; it shows an average error of about 1.5 kcal/mol for the prediction of absolute binding affinity of 37 and 128 known inhibitors of EphB4 and EGFR, respectively. High-throughput docking and ranking by two-parameter LIECE models are shown to be able to identify novel low-micromolar EphB4 and CDK2 inhibitors of low-molecular weight (< or =355 g/mol).