Abstract

Exploring predictive QSAR models for dopamine catechol structures could be used in designing more potent ligands. In this study, efforts were taken to find out the most important molecular features responsible for the biological activity of catechol structures. All 2D descriptors of Dragon including constitutional, topological, molecular walk counts, BCUT descriptors, Galvez topological, 2D autocorrelations, functional groups, atom-centred fragments, empirical descriptors and properties were calculated for the structures. Two non-linear modelling methods (PC-LS-SVM and PC-ANFIS) were used and compared in this QSAR study. The results revealed the more predictive ability of PC-LS-SVM in the QSAR analysis of the compounds with catechol substructure. The roles of topological properties and number of hydrogen bond donors group as molecular features responsible for the activity of the compounds were discussed. The obtained QSAR models can be used in future studies of drug development for human dopamine D2 receptor. © 2016, The Korean Society for Applied Biological Chemistry.