Quantitative relationships between structures of 26 diaryl substituted pyrazoles as CCR2 inhibitors and their activities were investigated by four linear and nonlinear methods namely MLR, PLS, GRNN and LS-SVM. The obtained models were able to describe about 83%, 87%, 86%, and 0.91% of the variance in the experimental activity of molecules in training set, respectively. The accuracy and predictability of the proposed models were illustrated using various evaluation techniques. Some of them were: cross-validation, validation through an external test set, and Y-randomization. Furthermore, various criteria suggested by Tropsha and Roy were applied for evaluation of predictability of developed models. A comparison between the four different developed methods indicates that LS-SVM can be preferred as supreme model.
Copyright (c) 2010 Elsevier Masson SAS. All rights reserved.