Vesna Rastija*, Maja Molnar, Tena Siladi and Vijay Hariram Masand
The aim of this study was to derive robust and reliable QSAR models for clarification and prediction of antioxidant activity of 43 heterocyclic and Schiff bases dipicolinic acid derivatives. Statistical performance of two different algorithms for splitting data into training and test set (randomly and ranking method), as well as models obtained by two set of descriptors calculated by different computer programs (DRAGON and ADMEWORKS). The best internal validation result has model obtained by DRAGON descriptors (MATS4m, EEig03d, BELm4, Mor10p), split by ranking method (R2 = 0.805; R2ext = 0.833; F = 30.914). The best external validation result has model obtained by ADMEWORKS descriptors (NDB, MATS5p, MDEN33, TPSA), split by random method (R2 = 0.692; R2ext = 0.848; F = 16.818). Important structural requirements for great antioxidant activity are: low number of double bonds in molecules; absence of tertial nitrogen atoms; higher number of hydrogen bond donors; enhanced molecular polarity; and symmetrical moiety. Two new compounds with potentially great antioxidant activities were proposed.
QSAR, antioxidant activity, dipicolinic acid, polar surface area, hydrogen bond donors.
Department of Chemistry, Faculty of Agriculture, J. J. Strossmayer University, Osijek 31 000, Department of Applied Chemistry and Ecology, Faculty of Food Technology, J. J. Strossmayer University, Osijek, Department of Applied Chemistry and Ecology, Faculty of Food Technology, J. J. Strossmayer University, Osijek, Department of Chemistry, Vidya Bharati College, Camp, Amravati, Maharashtra