Artificial intelligence models Traditional Chinese Medicine principles to integrate with biomedicine

Abstract

Using 26,339 model‐eligible Traditional Chinese Medicine (TCM) formula records, we trained an autoencoder following TCM's compression–expansion logic to learn the TCM embedding space (TCM‐ES), quantifying symptom‐pattern–formula mappings. TCM‐ES aligns with conventional TCM principles not used in training. Biomedical entities were mapped post hoc into the fixed space, supporting formula prediction, retrospective clinical analyses, and exploratory therapeutic prioritization.