ABSTRACT
Based on the analysis of information flow through the Artificial Neural Network (ANN), a new screening rule of the molecular structure parameter in Quantitative Structure-Activity Relationship (QSAR) was presented by comparing of the values of connection weights and biases of the ANN model. The result showed that model quality and prediction ability of QSAR model, which was constructed by screening structural parameter with ANN, were better than by the method with MLR. The method established the foundation for further study in the mechanism research of the bio-toxicity of organic chemicals.
ACKNOWLEDGMENT
The authors would like to thank the Harbin Institute of Technology for the financial support of this study.