Mehdi Hatami

The main purpose of the present study was modeling and prediction of the optical rotation ([M]D) of some biodegradable polymers containing α-amino acids using quantitative structure-activity relationship (QSAR) approaches. In order to attain this goal, the optical rotation of a collection of 53 polymers was selected as a data set. The data set was randomly divided into three sections, training, test and external validation sets. By using dragon software, various descriptors were calculated for all molecules in the data set. The important descriptors were selected applying genetic algorithm-partial least squares (GA-PLS) method. Then an artificial neural network (ANN) was written with MATLAB 7 and used these descriptors as inputs and its output was optical rotation of desired polymers. Then, the constructed network was used for the prediction of ([M]D values of validation set. The squared correlation coefficient R2 values of the ANN model for the training, test and validation sets were 0.998, 0.996 and 0.996 respectively. The results showed the ability of developed ANN to predict optical rotation of various polymers.