DEVELOPMENT OF NEURAL-NETWORK SIMULATOR FOR STRUCTURE-ACTIVITY CORRELATION OF MOLECULES (NECO) - PREDICTION OF ENDO EXO SUBSTITUTION OF NORBORNANE DERIVATIVES AND OF CARCINOGENIC ACTIVITY OF PAHS FROM C-13-NMRSHIFTS/

A perceptron type neural network simulator for structure-activity corr elation of molecules has been developed with two different learning me thods, i.e., back-propagation and reconstruction methods. First by use of the back-propagation method the exo/endo branching of norbornane a nd norbornene derivatives was correctly predicted from the set of C-13 NMR chemical shifts for various ring carbon atoms. Then the obtained correlation was analyzed by the reconstruction learning method. It was shown in this case that the MMR shifts for two carbon atoms out of se ven have strong correlation with the exo/endo branching. Further, stru cture-activity correlation between the C-13 NMR chemical shifts and ca rcinogenicity of 11 polycyclic aromatic hydrocarbons was also analyzed using the reconstruction method. It was demonstrated that neural netw ork analysis is suitable for the elucidation of complicated structure- activity problems where many factors are nonlinearly entangled.