Rational choice of bioactive conformations through use of conformation analysis and 3-way partial least squares modeling

Comparative molecular field analysis (CoMFA) has become widely used in three-dimensional (3D) QSAR studies. Although CoMFA has been of general use, there are some critical problems in the proper application. A major problem of CoMFA, including most other 3D QSAR methodologies, is that the results are dependent on the chosen bioactive conformations and the corresponding alignment rules of molecules. Recently, we have proposed a novel method with a 3-way PLS formulation for solving the conformation/alignment problem in 3D QSAR studies [K. Hasegawa, M. Arakawa, K. Funatsu, Chemom. Intell. Lab. Syst., 47 (1999) 33-40]. The purpose of the present study is to demonstrate the general utility of our approach by applying to a real CoMFA data set. The data set of Protein-Tyrosine Kinase (PTK) inhibitors was used as a test sample. The possible 3D conformations of all molecules were generated by conformational analysis and they were characterized by field variables of CoMFA. To each unique conformation of the most active compound, one sample-variable sheet comprising of the most similar conformations was defined. The 3-way arrays for 3-way PLS analysis were created by collecting all sample-variable sheets. From the regression coefficient values of the 3-way PLS model, conformations largely contributing to inhibitory activity were selected and the resulting final CoMFA model could give the reasonable 3D coefficient contour maps. (C) 2000 Elsevier Science B.V. All rights reserved.