QSAR Modeling of CCR5 Receptor Antagonists using Artificial Neural Network

Y.D. Aher and P. Garg (India)


Artificial intelligence, neural networks, QSAR, and CCR5


In-silico prediction methods are gaining the popularity in drug discovery processes as they are relatively inexpensive and less time consuming. In this study, Artificial Neural Network (ANN) based on back propagation algorithm (BP algorithm) has been applied to develop a Quantitative structure-activity relationship (QSAR) model for predicting pIC50 values of a series of structurally diverse CCR5 receptor antagonists. Model was trained with a data set of 426 chemical compounds and tested with independent data set of 423 compounds. A good correlation between training data set (R2 = 0.847) and total data set of 849 chemical compounds (R2 =0.765) is observed. The results are further validated by performing external validation using validation set (R2 =0.692) and leave 20% out cross validation which show that such an approach is suitable for establishing a predictive model to determine pIC50 values.

Important Links:

Go Back