3rd Mediterranean Symposium on Medicinal and Aromatic Plants (MESMAP-3), Girne, Cyprus (Kktc), 13 - 16 April 2017, pp.221-275
In this work, we presents the results of pharmacophore identification and bioactivity prediction for phenylprazine analogues as class of PDE10A inhibitors by means of Electron Conformational-Genetic Algorithm method1. EC-GA method was developed that combines EC and GA methods. 3D structures of compounds, conformational analysis and quantum chemical calculations have been worked Spartan 10 software at Hartree Fock 3-21 G* level and water was used as solvent. Then the ECMC matrix was prepared that contains distances between atoms and mulliken charges for pharmacophore identification. Nonlinear least square regression method and genetic algorithm were performed to predict the theoretical activity. For cross validation and statistical analysis compouds were classified training and test set. In this study 49 training, 24 test set was used. For the predictive capabilities best subset of descriptors were calculated for a range of 1-14 parameters at MATLAB software. Hence the model attains a stable situation as optimum 10 parameter number. LOO-CV method had been employed for the predictive ability of 4D-QSAR model. In addition to E-statistical method was used to determination of the effect selected parameters. R2training, R2test, q2, q2ext1, q2ext2, q2ext3 and con1, con2, con3 values are given respectively 0.865, 0.819, 0.808, 0.824, 0.816, 0.819, 0.929,0.899, 0.920.
Keywords: 4D-QSAR, EC-GA, Phenylpyrazine, Pharmacophore