(Received: January 10, 2002; Accepted for publication: February 12, 2002; Published on Web: March 22, 2002)
A neural network was applied to the prediction of the carcinogenicity of 41 kinds of organic chlorine-containing compounds. Seven kinds of structural and quantum-chemical descriptors: molecular weight, log P, Gibbs free energy, ionization potential, LUMO energy, HOMO-LUMO energy gap, and Connolly volume were determined. These descriptors were entered into the input layer of a three-layered neural network, and carcinogenicity data from the NTP database were entered into the output layer as teaching data. The network was trained with an error-back-propagation method, and a leave-one-out test showed a correct classification rate of 93%.
Keywords: Structure-activity relationship, Neural network, Carcinogenicity prediction, Chlorine-containing organic compounds
Text in Japanese