Analytical Sciences

Prediction of Acute in vivo Toxicity of Some Amine and Amide Drugs to Rats by Multiple Linear Regression, Partial Least Squares and an Artificial Neural Network

Mohamad Khayatzadeh MAHANI,* Marzieh CHALOOSI,* Mohamad Ghanadi MARAGHEH,** Ali Reza KHANCHI,** and Daryoush AFZALI***

*Faculty of Chemistry, Tarbiat Moalem University, 49 Mofateh Avenue, Tehran, Iran
**Jaber Ibn Hayan Research Laboratories, Nuclear Science and Technology Research Institute, AEOI, P. O. Box 11365-8486, Tehran, Iran
***Faculty of Science, Department of Chemistry, Shahid Bahonar University, Kerman, Iran

The oral acute in vivo toxicity of 32 amine and amide drugs was related to their structural-dependent properties. Genetic algorithm-partial least-squares and stepwise variable selection was applied to select of meaningful descriptors. Multiple linear regression (MLR), artificial neural network (ANN) and partial least square (PLS) models were created with selected descriptors. The predictive ability of all three models was evaluated and compared on a set of five drugs, which were not used in modeling steps. Average errors of 0.168, 0.169 and 0.259 were obtained for MLR, ANN and PLS, respectively.

J-STAGE: View this article in J-STAGE