NEW DRUG R&D OF TRADITIONAL CHINESE MEDICINE: ROLE OF DATA MINING APPROACHES

Abstract

Traditional Chinese Medicine (TCM) documented about 100,000 formulae during past 2500 years. To use and customize them by modern pharmaceutical industry, we make an interdisciplinary effort to study the activity of new drug research and development (R&D) in TCM by introducing data mining approaches to it. We used the migraine formulae as a training set to investigate the possibility of developing new prescription by means of data mining. The activity of new drug R&D of TCM consists of two steps. The first step is to discover new prescriptions (candidates for drugs) from migraine formulae. We present an unsupervised clustering approach based on data mining theory to address the problem in the first step and automatically discover ten new prescriptions from the formulae data. The second step is to develop and optimize the prescriptions discovered by current biomedical approaches. Since Ligusticum chuanxiong Hort (LCH), a kind of herb, is often used to treat migraine and appears in the new prescriptions, we use it as an example and apply supervised regression method based on data mining theory to study the drug R&D activity of TCM. We revised two linear regression methods in order to establish the nonlinear association between three chemical ingredients of LCH and corresponding pharmacological activity and used it to predict the activities. The association is validated by in vitro experiments and we found that the experimental results are consistent with the prediction. Unsupervised clustering and supervised regression cover most part of data mining theory, which means that data mining approaches play a crucial role in new drug R&D in TCM and present a better solution to establish the platform of drug R&D in TCM.