Dévelopement d'un Modéle Qsar Avancé pour la Prédiction d'ctivités Ames de Molécules Nitro.aromatique

Abstract

The work presented in this ains to develop reliable stable and predictive OSAR models for predicting logat 100 properties of nitroaromatic compounds Two types of OSAR models .:regression and classification were used to model the toxicity of substances represented by the log AT 100 of nitroaromatic compounds the regression model with an R2 OF 72.96 and a q4 of 71.58 was corrzlated with four molecular descriptors RCL 0SM2 DZ ATSCLm and MaxddsN Thes results in accordance with the main recommendation of Golbraikh and Tropshathe machine leaming classification model .logistic Regression LRegression achived a sensitivity of 85.71 a specificity of 80.43 and an accuracy of 76.92 These result indicate that this model has excellent internal predictive parameters as well as robustness and stability.