Discovering New Drug in Ancient Herbal Compound Database by Unsupervised Pattern Discovery Algorithm

Combining advanced data mining and biomedical technologies to discovering new drug is an active research field nowadays. In this paper, we collect a herbal compounds for rheum database by searching about 150 prescriptions in ancient herbal document. 255 herbal compounds are included for their combinations to heal rheum. Our aim is to discover potentially new herbal compound in the database. We present the unsupervised pattern discovery algorithm to allocate the herbal compounds into different cluster in a self-organized way and obtain 42 clusters, some of which fully accord with Chinese medicine theory and the other can be considered as the potential new drug, which need to be validated by pharmacology further. We also present an executable and effect strategy for further experiments. We conclude that data mining methods, especially, unsupervised learning method, can be taken as a new technique to discovering new drugs.