Abstract:

 A procedure is presented for the prediction of physical property of organic compound and QSAR/QSPR analysis based on the similarity indices using a back-propagation neural network. The similarity indices were calculated on a chosen set of structural descriptors by equation 1 and used to quantify the similarity or dissimilarity of organic compound. The similarity indices were also used as an input parameter of neural network. The pKa values of the 69 substituted phenols were predicted by using 32 compounds as a training set and all 69 compounds as a predicting set. The results obtained were satisfying.

Key words: Molecular similarity, Phenols, Neural network