अमूर्त

Quantitative structure - Activity relationships study of carbonic anhydrase inhibitors using multinomial logistic regression model and artificial neural networks

Hassan Sahebjamee, Parviz Abdolmaleki, Alireza Foroumadi, Parichehre Yaghmaei

Multinomial logistic regression (MLR) and artificial neural networks (ANNs) were employed to seek the quantitative structure – activity relationships (QSARs) that correlate structural descriptors and inhibition activity of carbonic anhydrase IX inhibitors. Many quantitative descriptors (n=644) were generated to express the physicochemical properties of 132 compounds with optimized structures with known ki values.MLR were used to nonlinearly select different subsets of descriptors and develop nonlinear models for prediction of log (ki). The most significant parameters were then selected. A neural network model was then constructed and fed by the parameters selected by MLR. The networks have been trained and tested using the best subset selected by MLR. The best prediction model was found to be a 5-3-3 artificial neural network which was fed by the most frequently selected descriptors among these subsets. Cross-validation and a separate prediction set were used to evaluate the stability and prediction ability of the establishedmodels. Our results demonstrated that descriptors correlated to autocorrelations, topological properties weremajor determinants of inhibition activity of these compounds. Bothmethods were able to significantly describe and predict the CAIX inhibitory activity.