Topological QSAR modeling of cytotoxicity data of anti-HIV 5-phenyl-l-phenylamino-imidazole derivatives using GFA, G/PLS, FA and PCRA techniques
Kunal Roy, Joseph Thomas Leonard · INDIAN JOURNAL OF CHEMISTRY- SECTION A · 2006
Cytotoxicity data of anti-HIV 5-phenyl-l-phenylamino-imidazole derivatives have been subjected to QSAR study using topological and structural descriptors applying techniques like stepwise regression, multiple linear regression with genetic function approximation, genetic partial least squares, multiple linear regression with factor analysis as the preprocessing step and principal component regression analysis. Presence of hydrogen bond donor groups appear to be an important feature for reducing the cytotoxicity. Molecular size, shape and branching are also found important for determining the cytotoxicity. Different fragments of the molecules have been identified as significant contributors in modulating the cytotoxicity. Among all parameters used, H_bond_donor and E-state parameters are most prevalent in different models. The quality of equations obtained from the different techniques are in the comparable range (explained variance ranging from 63.9-68.8% while predicted variance ranging from 61.5-64.5%). However, the best equation statistics were obtained with principal component regression analysis (explained variance and predicted variance being 81.0% and 75.8%, respectively).