AI tools help decode how TCM formulas work

Traditional Chinese medicine formulas (TCMFs) are widely used in clinical practice, but their molecular mechanisms can be difficult to pin down because formulas often contain many constituents that act on multiple targets and pathways. In a new review in the Chinese Journal of Natural Medicines, researchers describe how artificial intelligence (AI) methods are increasingly being used to tackle this complexity and accelerate mechanism-oriented research on TCMFs.

The review, authored by researchers at Shanghai University of Traditional Chinese Medicine and collaborators, surveys AI-enabled strategies across several key steps in “from formula to mechanism” workflows, including predicting pharmacokinetic behavior, narrowing down candidate targets, identifying synergistic interactions, and guiding optimization of herbal combinations. TCMFs are grounded in holistic theory and long-term clinical experience, but modern mechanistic studies face practical barriers. A single formula may contain dozens to hundreds of chemical constituents, with effects that can vary by absorption, distribution, metabolism, and excretion (ADME), as well as by biological context. These factors create a high-dimensional search space where conventional one-compound/one-target approaches may be insufficient.

According to the authors, AI applications in TCMF molecular mechanism research are expanding in several directions: Machine learning models can help estimate pharmacokinetic properties for compounds in herbal medicines, supporting prioritization of constituents more likely to be bioavailable and relevant in vivo. AI-assisted modeling can support target prediction, helping connect candidate constituents to putative protein targets or pathways for downstream validation. Because formulas are designed around compatibility and combination principles, AI approaches are being explored to detect and interpret compound-compound and compound-target synergy patterns that may underlie observed therapeutic effects. Computational approaches can help integrate multi-source data and generate testable hypotheses about pharmacological mechanisms, supporting iterative cycles of prediction and experimental confirmation.

The review discusses how AI may assist in optimizing herbal combinations, potentially improving efficacy, safety, or consistency by guiding selection and balancing of components. The authors also highlight challenges and opportunities for the field, positioning AI as a driver for the modernization and globalization of traditional Chinese medicine research, provided that methodological rigor, data quality, and appropriate validation frameworks keep pace with rapid tool development.