Abstract
In order to study the structure-activity relationship between 13C-NMR chemical shifts and the carcinogenicity of eleven polycyclic aromatic hydrocarbons (PAH), a neural network analysis was performed with the reconstruction method which is a pre-learning method and the relationship between input and output data is suggested.
The neural network applied is a standard perceptron type with three layers. In this analysis 13C-NMR chemical shifts are treated as structuraldata and the carcinogenicity as output.
It has been reported that electron density in three regions, i.e., L, K and bay regions has strong correlation with the carcinogenicity of PAH.
Importance of these three regions is suggested by reconstructing the weight matrices.