Author(s)

H. Shiroyama, Y. Zuo & E. Kita

Abstract

In drug discovery, it is very important to predict the side effect of the drug accurately. The prediction algorithm of the drug side effect is presented in this study. This algorithm is based on the concept of the structure-activity relationship. Firstly, the drug side effects are gathered from the registration of medical products by using text mining. Next, the chemical structure information of the drug is obtained from the PubChem data base. Then, the association rules between the chemical structure and the side effects are defined. The associate rules are applied to the prediction of the side effect of 10 chemical products. Keywords: drug, side effect, association rule, PubChem, text mining. 1 Introduction Several drugs (medicines) have been developed every year. While new drugs are very useful for improving illness and injuries, they sometimes have terrible side effects. Therefore, it is very important for the prediction of the drug side effects in the drug discovery. A new drug discovery is a very time-consuming process. The drug discovery is mainly composed of four steps; basic study, non-medical study, medical study and approval and production. In the basic study, the potential chemical products are developed. In non-medical study, the effect of the products is confirmed in animal experiment and so on. In medical study, the effect of the products is provided for patients and health persons. Since the side effects of the potential products are confirmed in non-medical and medical studies, the drug discovery needs a long time and enormous cost. Therefore, some researchers have studied the prediction algorithm of the drug side effect before non-medical and medical studies. Enslein et al. used

Keywords

drug, side effect, association rule, PubChem, text mining