Traditional Chinese medicine (TCM) has long been used in the clinical treatment of coronary heart disease (CHD). TCM is characterized by syndrome-based medication, which is, using different TCM formulae for different syndromes. However, the underlying mode of action remains unclear. In this work, we utilized network pharmacology and machine learning to explore the mechanism of eight classic TCM formulae in the treatment of different types of CHD. First, by integrating multiple databases, a total of 669 potential bioactive compounds and 581 targets of the eight formulae were screened. Then, the effectiveness of these formulae on CHD was evaluated using two network-based indicators. The results showed that each formula's targets were significantly correlated with CHD associated genes and overlapped with the targets of 9 classes of drugs for cardio vascular diseases (CVD) to some degree. Next, from 5 different levels, i.e., herb, symptom, compound, target, and pathway level, we systematically compared the eight formulae using network clustering and hierarchical clustering. We found that all the formulae could be grouped into five clusters and the clustering results were approximately consistent at different levels. All the formulae were involved in 7 pathways closely related to CHD and may exhibit the common effect of relieving angina. Formulae in the same group collectively regulated some unique pathways and suggest further specific indications. For example, the three formulae used for Qi stagnation and blood stasis, Qi deficiency and blood stasis, and Qi-Yin deficiency syndromes acted on two special pathways (TNF signaling pathway, NF-kappa B signaling pathway) and may exert anti-inflammatory and immune-enhancing effects; the two formulae for Yin deficiency of heart and kidney, and Yang deficiency of heart and kidney syndromes regulated two special pathways (PPAR signaling pathway, thyroid hormone signaling pathway) in endocrine system and could improve renal function. Subsequently, we designed a rank algorithm, which integrated network topology with biological function, to identify important targets of these formulae. The results were consistent with the multi-level clustering results. At last, our literature mining validated about 20 % putative targets, as well as clustering results and effects of the formulae by experimental evidences. This study explained the medication patterns and scientific significance of TCM formulae on different types of CHD from perspective of systems biology. It may facilitate the understanding of different types of CHD described by traditional Chinese medicine from the perspectives of modern biology.
Keywords: Coronary heart disease; Machine learning; Modern medicine symptom; Network pharmacology; TCM syndrome; Traditional Chinese medicine.
Copyright © 2020. Published by Elsevier Ltd.