A Knowledge Graph-Driven Hypergeometric Efficacy Prediction Model for Classical Traditional Chinese Herbal Formulas.

Area of Science:

• Computational biology
• Pharmacology
• Traditional Chinese Medicine

Background:

• Traditional Chinese Medicine (TCM) formulas possess complex semantic structures hindering computational representation of their efficacy.
• Existing methods struggle to represent the multi-level efficacy of TCM formulas in a computable format.

Purpose of the Study:

• To develop a statistically rigorous and interpretable computational model for predicting the dominant efficacies of classical TCM herbal formulas.
• To enhance the computability and integrity of TCM efficacy information.

Main Methods:

• Constructed a knowledge graph with five semantic entities: disease, syndrome, symptom, efficacy, and herb.
• Developed the Hypergeometric Efficacy Prediction Model (HEPM) using hypergeometric enrichment analysis.
• Validated the model on a curated dataset of 174 classical TCM formulas.

Main Results:

• HEPM accurately reproduced characteristic efficacy patterns of classical prescriptions with an average F1-score of 0.63.
• The knowledge graph resolved semantic inconsistencies and incompleteness in traditional efficacy descriptions.
• The model demonstrated enhanced integrity and computability of TCM efficacy information.

Conclusions:

• HEPM offers a statistically grounded and interpretable framework for modeling TCM formula efficacy.
• This method provides a replicable approach for efficacy prediction in TCM.
• The framework supports knowledge-driven intelligent TCM analysis and clinical decision-support systems.