Identifying Active Compounds from Chinese Medicinal Plants via Causal Variable Selection

Medicinal plants are growing to be a major source of drug discovery, and one challenging problem is to identify the drug candidates from numerous ingredients in medicinal plants. We present an approach to identify active compounds from Chinese medicinal plant based on causal variable selection techniques. We examined three methods, including stepwise regression, Incremental Association Markov Blanket (IAMB) and Grow-Shrink Markov Blanket (GS), in term of their ability to generate robust results. IAMB outperformed the other two methods and was applied to identify active compounds from one well-known Chinese medicinal prescription widely used to treat cardiovascular diseases. The biological relevance of identified compounds was confirmed by literature knowledge and experimental data, and thereby suggested our method be promising to facilitate the drug discovery from medicinal plants.