Models for predicting IKKA and IKKB blockade.

We describe the application of different methods in the development of QSAR models for IKKA and IKKB inhibition. The results show that the best QSAR models provide highly accurate predictions for existing IkB-kinase (IKK) inhibitors. The exceptions, corresponding to 5% of the known collection of inhibitors, are five classes of compounds incorporating the nitrile or sulfonamide moieties, small compounds with molecular weights of less than 300, and two classes of blockers considered to be type II kinase inhibitors. Comparison of our novel IKKB homology model and the recently reported IKKB crystal structure implies that a predictive protein-antagonist complex structure is more likely to exist as an inactive form in the crystalline state as observed in the recent protein X-ray structure.