QSAR Study of Arylsulfonylpiperazine Inhibitors of 11β-HSD1 by GA-MLR, GA-PLS and GA-ANN

Abstract

The main goal of this study is to establish robust theoretical models for prediction of pIC values related to arylsulfonylpiperazines. After drawing the molecular structures, a suitable set of molecular descriptors that fulfill the best fitted models were selected using genetic algorithm. Three models were proposed for prediction of experimental results involving multiple linear regression (MLR), partial least square (PLS) and artificial neural network (ANN) techniques. The accuracy of the suggested models was confirmed by cross-validation, validation through an external test set, and Y-randomization. Numerical values predicted by all the three models are in good agreement with the experimental ones. However, the results derived from ANN technique show better compatibility (R² =0.930, R2 =0.788, RMSE =0.137, RMSE =0.320, Ftrain=72.850, Ftest=1.743) with the actual experimental values.

Keywords:

• QSAR
• Arylsulfonylpiperazine
• 11β-HSD1
• MLR
• PLS
• ANN