Prediction of Thrombin and Factor Xa Inhibitory Activity with Associative Neural Networks

Quantitative structure-activity relationship studies on a series of selective inhibitors of thrombin and factor Xa were performed by using Associative Neural Network. To overcome the problem of overfitting due to descriptor selection, 5-fold cross-validation with variable selection in each step of the analysis was performed. The predictive ability of the models was tested through leave-one-out cross-validation, giving a Q 2 =0.74 - 0.87 for regression models. Predictions for the external evaluation sets obtained accuracies in the range of 0.71 - 0.82 for regressions. The proposed models can be potential tools for finding new drug candidates.