Prediction chemical toxicity of organic pollutants using artificial neural networks (ANN)Author(s): Mehdi Alizadeh
An artificial neural networks (ANN) study, has been carried out on 38 diverse organic pollutants for prediction chemical toxicity by usingmolecular structural descriptors. Modeling of logarithm values of chronic toxicity in fish (Log Chv) of these compounds as a function of the theoretically derived descriptors was established by artificial neural networks (ANN). The Stepwise SPSS was used for the selection of the variables (descriptors) that resulted in the best-fitted models. For prediction Log Chv of compounds, three descriptors were used to develop a quantitative relationship between the Log Chv and structural properties. Appropriate models with low standard errors and high correlation coefficients were obtained.After variables selection, compounds randomly were divided into two training and test sets and ANN used for building the best models. The predictive quality of theANNmodels were tested for an external prediction set of 11 compounds randomly chosen from 38 compounds. The regression coefficients of prediction for the ANN model were 0.9940, 0.9955 for training and test sets respectively. Result obtained showed that ANN can simulate the relationship between structural descriptors and the Log Chv of the molecules in data sets accurately.