Prediction of progesterone receptor inhibition b y high"performance neural network algorithm
I. V. Romero Reyes, Irina V. Fedyushkina, Vladlen S. Skvortsov, Dmitry A. Filimonov · 2013
The computing model based on data of the interaction of progesterone receptor inhibitors with ligand binding domain was developed. It allowed estimating inhibition constant for probable inhibitors with high accuracy ( 2 0.95 R = ). This model is based on t he feed forward back propagation neural network with the LevenbergMarquardt method and the GPU parallel program realization was implemented. The last one speeded up in 734 times as compared with serial MATLAB realization. The parallel realization significantly increased the size of the processed data, while reducing timeconsuming.