OBJECTIVE To identify compound combinations as candidate multi-component drugs for the type 2 diabetes from natural product information.METHODS Chemical composition information of herbs in natural medicine was acquired by integrating conventional databases;Traditional Chinese Medicine Information Database(TCM-ID)and Traditional Chinese Medicine Integrated Database(TCMID).Therapeutic effect of each herb on the type 2 diabetes was examined by analyzing annotated function information with a text-mining method.The Apriori algorithm,which is a classical method for extracting associations between object in large-scale databases,was employed to infer association rules between compound combinations and therapeutic effect on the target disease.The chemical composition and therapeutic information of each herb was used as a transaction,which consists of the chemical compound combination as an antecedent item set and the therapeutic effect as a consequent item.The association rules with high support and confidence value were suggested as candidate multi-component drugs for the type 2 diabetes.RESULTS Totally 40 941 association rules were inferred with support lower bound 0.05% and maximum rule length 4.With respect to support and confidence,the top-ranked compound combination was puerarin and daidzin(support=0.15%,confidence=100%).In addition,the top 16 compound combinations were composed of 11 individual chemical compounds;puerarin,daidzin,abscisic acid,batatisine,dopamine,cholesterol,daidzein,gamma-aminobutyric acid,stigmasterol,campesteryl ferulate,and campesterol.To validate therapeutic effect of the proposed compound combinations,literature evidences of each individual compound were investigated.Among the 11 individual compounds,six compounds were reported to be effective for the treatment of the diabetes mellitus.CONCLUSION By analyzing natural product in formation with association rule mining,16 compound combinations are suggested as candidate multi-component drugs for the type 2 diabetes.These compound combinations are recommended for further investigation in the context of drug development. OBJECTIVE To identify compound combinations as candidate multi-component drugs for the type 2 diabetes from natural product information. METHODS Chemical composition information of herbs in natural medicine was acquired by incorporated conventional databases; Traditional Chinese Medicine Information Database (TCM-ID) and Traditional Chinese Medicine Integrated Database (TCMID). Therapeutic effect of each herb on the type 2 diabetes was examined by analyzing annotated function information with a text-mining method. The Apriori algorithm, which is a classical method for disordering associations between object in large-scale databases , was employed to infer association rules between compound combinations and therapeutic effects on the target disease. The chemical composition and therapeutic information of each herb was used as a transaction, which consists of the chemical compound combination as an antecedent item set and the therapeutic effect as a consequent item. the association rules with high suppo rt and confidence value were suggested as candidate multi-component drugs for the type 2 diabetes .RESULTS Totally 40 941 association rules were inferred with support lower bound 0.05% and maximum rule length 4.With respect to support and confidence, the top-ranked compound combination was puerarin and daidzin (support = 0.15%, confidence = 100%). In addition, the top 16 compound combinations are composed of 11 individual chemical compounds; puerarin, daidzin, abscisic acid, batatisine, dopamine, cholesterol, daidzein, aminobutyric acid, stigmasterol, campesteryl ferulate, and campesterol. To validate therapeutic effect of the proposed compound combinations, literature evidences of each individual compound were investigated. Among the 11 individual compounds, six compounds were reported for be effective for the treatment of the diabetes mellitus .CONCLUSION By analyzing natural product in formation with association rule mining, 16 compound combinations are suggested as candidate multi-component drugs for the type 2 diabetes.These compound combinations are recommended for further investigation in the context of drug development.