ABSTRACT

2D QSAR has been performed on a series of pyridine carbonitrile and trifluoromethyl phenyl derivatives. 53 compounds were divided into training and test sets out of which 37 compounds generated a final QSAR model. The most significant model with n = 37, r = 0.916, r2 = 0.762, r2 cv = 0.759, s value = 0.388, f value = 41.76 was developed using MLR analysis. For PLS, the fraction of variance explained = 0.806 was observed. A comparable PLS model with r2 = 0.806 and Neural model with r2 = 0.853 indicated good internal predictability of the model. External test set validation provided r2 values of 0.744 and 0.768 for MLR and PLS analysis, respectively. Dipole moment Z Component, Log P, Shape flexibility index, and Vamp LUMO descriptors proved to be significant for inhibition of Cathepsin S. These findings will be effective in designing more potent and effective Cathepsin S inhibitors