Prediction of Carcinogenicity of Organic Chlorine-containing Compounds by Neural Network

概要

論文の詳細を見る
In order to develop a chemical hazard prediction system showing a higher performance than those of existing systems using linear multi-regression analysis as quantitative structure-activity relationship, neural network was applied to the prediction of carcinogenicity of 41 kinds of organic chlorine-containing compounds. Seven kinds of structural descriptors easily determined from molecular structures were used. A leave-one-out test showed a performance of 83%, higher than those of existing systems, and a possibility to get a higher performance was discussed.
日本コンピュータ化学会の論文