A Study on QSAR of Environmental Pollution Based on Back-Propagation Neural Networks

Experimental research shows that the literature data on the direct-acting mutagenicities in the Salmonella Typhimurium strain TA98 of nitrated Polycyclic Aromatic Hydrocarbons(PAHs), together with a descriptor represented in this paper, are approximately accordant with the result of a systematic analysis of the performance of back-propagation neural-network models. Based upon two classes of activities, the correct recognition rate can reach up to 97 persent, which is better than the rate obtained through the use of pattern recognition. These results can lead to general recommendations for a proper evaluation of recognition and prediction criteria for this class of nonlinear structure-activity models.