薬物相互作用予測へのニューラルネットワークの応用(<一般演題>複合領域)

抄録

I have investigated for the predicting performance of drug interactions including drug properties which influence pharmacologic actions of other drugs using layered neural networks. Drug interactions mean that the pharmacologic action is different from the case of the monotherapy in a drug for reason why interactions of something exist between some drugs. In this study, I considered the drug metabolism in liver, the induction or reduction of hepatic metabolic enzymes, the kidney function, the influence of the renal failure, the absorption in stomach and the neutralization of stomach acid as drug properties. The neural networks using this study includes two subnetworks, one of the subnetwork images drug properties which influence pharmacologic actions, the other one predicts the interaction grade including the decrease, increase or no change of pharmacologic effects recognized as drug interaction. The back propagation algorithm was used as a learning method of these neural networks. The drug interaction informations of real drugs in Japan were used as learning data. I evaluated the predicting performance using these data and neural networks, and considered for the difference in the predicting power depending on learning data counts.