Predicting Meridian in Chinese traditional medicine using machine learning approaches

Abstract

Plant-derived nature products, known as herb formulas, have been commonly used in Traditional Chinese Medicine (TCM) for disease prevention and treatment. The herbs have been traditionally classified into different categories according to the TCM Organ systems known as Meridians. Despite the increasing knowledge on the active components of the herbs, the rationale of Meridian classification remains poorly understood. In this study, we took a machine learning approach to explore the classification of Meridian. We determined the molecule features for 646 herbs and their active components including structure-based fingerprints and ADME properties (absorption, distribution, metabolism and excretion), and found that the Meridian can be predicted by machine learning approaches with a top accuracy of 0.83. We also identified the top compound features that were important for the Meridian prediction. To the best of our knowledge, this is the first time that molecular properties of the herb compounds are associated with the TCM Meridians. Taken together, the machine learning approach may provide novel insights for the understanding of molecular evidence of Meridians in TCM.

Anticancer Plants: A Review of the Active Phytochemicals, Applications in Animal Models, and Regulatory Aspects

“…In addition, several researchers have developed strategies for in silico pharmacokinetic properties of molecules/drugs [187][188][189][190][191]. Such approaches are also applicable to phytochemicals and plant-based active drug components for their virtual screening, possible mode of action, and advanced drug discovery [192][193][194][195]. Several plant-based anticancer compounds have been evaluated using in silico and systems pharmacology tools [196][197][198][199][200][201].…”

Predicting the Associations between Meridians and Chinese Traditional Medicine Using a Cost-Sensitive Graph Convolutional Neural Network

“…Previous works took the average of the compound fingerprint features in the herb, and applied random forest algorithm to classify seven major Meridians including the lung, liver, stomach, spleen, kidney, heart, and large intestine [15]. In deep learning approach, previous studies learned node-level representation using GCN model for further classification [23,24].…”

“…Methods ROC-AUC Average fingerprint Random forest [15] 0.65 Neural fingerprint GCN [23,24] 0.70 Neural fingerprint Cost-sensitive GCN 0.78…”

“…Many researchers paid attentions on TCM as a complementary and alternative medicine with Western medicine [10][11][12]. With the increasing knowledge about the chemical ingredients of TCM herbs, there are strong needs to provide a systematic strategy for identifying the Meridians in TCM through ingredient compounds [13][14][15][16].…”

Unsupervised Learning and Multipartite Network Models: A Promising Approach for Understanding Traditional Medicine

“…To achieve the ultimate goal of precision medicine, i.e., the right intervention for a patient at the right time ( Stefano and Kream, 2015 ), there has been a long history of symptom-based diagnosis that utilizes available information to classify patients, diseases, and drugs ( Figure 1 ). In the early days of traditional medicine, physicians tried to characterize diseases using empirical terms, such as temperament and meridian ( Rezadoost et al., 2016 ; Arji et al., 2019 ; Wang Y. et al., 2019 ), based on which they prescribed corresponding herbs that are known to target them ( Xu, 2011 ; Li and Weng, 2017 ). With increasing knowledge on biochemistry, the era of modern medicine has started, further advancing our understanding of human diseases to the molecular level.…”