Comparative molecular field analysis of flavonoid inhibitors of the PIM-1 kinase

A predictive CoMFA model of flavonoid inhibitors of the PIM-1 kinase is reported. Shown is the alignment rule for the training set of compounds used to build the CoMFA model. The PIM-1 protein, the product of the pim-1 oncogene, is a serine/threonine kinase. Dysregulation of the PIM-1 kinase has been implicated in the development of human malignancies including lymphomas, leukemias, and prostate cancer. Comparative molecular field analysis (CoMFA) is a 3-D QSAR technique that has been widely used, with notable success, to correlate biological activity with the steric and electrostatic properties of ligands. We have used a set of 15 flavonoid inhibitors of the PIM-1 kinase, aligned de novo by common substructure, to generate a CoMFA model for the purpose of elucidating the steric and electrostatic properties involved in flavonoid binding to the PIM-1 kinase. Partial least squares correlation between observed and predicted inhibitor potency (expressed as −log IC 50), using a non-cross-validated partial least squares analysis, generated a non-cross-validated q 2 = 0.805 for the training set ( n = 15) of flavonoids. The CoMFA generated steric map indicated that the PIM-1-binding site was sterically hindered, leading to more efficient binding of planar molecules over ( R) or ( S) compounds. The electrostatic map identified that positive charges near the flavonoid atom C8 and negative charges near C4′ increased flavonoid binding. The CoMFA model accurately predicted the potency of a test set of flavonoids ( n = 6), generating a correlation between observed and predicted potency of q 2 = 0.825. CoMFA models generated from additional alignment rules, which were guided by co-crystal structure ligand orientations, did not improve the correlative value of the model. Superimposing the PIM-1 kinase crystal structure onto the CoMFA contours validated the steric and electrostatic maps, elucidating the amino acid residues that potentially contribute to the CoMFA fields. Thus we have generated the first predictive model that may be used for the rational design of small-molecule inhibitors of the PIM-1 kinase.