Quantitative Structure Activity Relationship (QSAR) modeling of opioid activities for some morphine derivatives by linear non-linear procedures

QSAR studies have been performed on forty-four molecules of morphine derivatives. With studies on opioid activities for some morphine derivatives [1] and determining the amount of agonist potency relative to morphine [2], tow techniques: multiple linear regression (MLR) and support vector machine (SVM) were used to design the relationship between molecular descriptor and agonist potency morphine derivatives (14-Acylaminomorphinones and 14-alkylaminomorphinones). Semi- empirical quantum chemical calculation (AM1 method) was used to fine the optimum 3D geometry of the studied molecules. Among these 44 molecules, 35 (79%) were selected as the training set molecules based on structural diversity and the remaining 9 (21%) molecules were used as test set. The modeling was estimated by Stepwise selection method using different descriptors obtained from Dragon software. Four hundred thirty-six descriptors were calculated. The coefficient of determination (R-square) of 0.9 was observed between experimental and predicted activity value of training set and R-square of 0.821 was observed for test set by MLR. The coefficient of determination of 0.939 was observed for training set and R-square of 0.863 was observed for test set by SVM. The standard error was calculated 0.344 for training set and 0.364 for test set by MLR and it was calculated 0.077 for training set and 0.218 for test set by SVM. High correlation between experimental and predicted activity values was observed, indicating the validation and the good quality of the derived QSAR models