Meridian-GAT: Modeling Meridian System Mechanisms Using Graph Attention Networks
DOI:
https://doi.org/10.71222/vha7wc31Keywords:
meridian system, graph neural networks, graph attention network, acupuncture mechanism, complex networks, traditional Chinese medicineAbstract
The meridian system is a central theoretical component of traditional Chinese medicine (TCM), describing functional pathways through which acupuncture stimulation is transmitted to regulate physiological states. Despite its extensive clinical use, the meridian system lacks a unified computational framework capable of quantitatively modeling its network structure and transmission mechanisms. In this study, we formulate the meridian system as a complex graph, where acupoints are represented as nodes and meridian-based and functional relationships are represented as edges, and propose Meridian-GAT, a graph attention network-based model for modeling meridian system mechanisms. By leveraging attention mechanisms in graph neural networks, the proposed model captures heterogeneous and non-uniform transmission strengths among acupoints, enabling data-driven exploration of meridian connectivity and information propagation patterns. Multi-dimensional acupoint features, including spatial attributes, meridian affiliations, and functional indications, are integrated into a unified graph representation to support mechanism-oriented learning tasks. Meridian-GAT is evaluated on acupoint representation learning and acupuncture efficacy prediction tasks using a curated meridian knowledge dataset. Experimental results demonstrate that Meridian-GAT outperforms baseline graph neural network models, achieving an improvement of 8.7% in prediction accuracy compared with the standard GCN model. Furthermore, the learned attention weights provide interpretable insights into key acupoints and dominant transmission pathways, which are consistent with classical meridian theory. This work offers a novel graph-based computational framework for quantitatively modeling meridian system mechanisms and contributes to the scientific interpretation and modernization of acupuncture theory.References
1. H. M. Langevin, and J. A. Yandow, "Relationship of acupuncture points and meridians to connective tissue planes," The Anatomical Record: An Official Publication of the American Association of Anatomists, vol. 269, no. 6, pp. 257-265, 2002. doi: 10.1002/ar.10185
2. A. C. Ahn, A. P. Colbert, B. J. Anderson, G. Martinsen, R. Hammerschlag, S. Cina, and H. M. Langevin, "Electrical properties of acupuncture points and meridians: a systematic review," Bioelectromagnetics: Journal of the Bioelectromagnetics Society, The Society for Physical Regulation in Biology and Medicine, The European Bioelectromagnetics Association, vol. 29, no. 4, pp. 245-256, 2008.
3. Y. Liu, and Y. Zhang, "Research on the association between acupuncture points and diseases based on the construction of knowledge graph and applied research," J. COMBIN. MATH. COMBIN. COMPUT, vol. 127, pp. 3355-3373, 2025.
4. L. I. U. Wei, W. A. N. G. Lan, F. A. N. Zhi-xin, K. A. N. G. Yi-zhe, Y. A. O. Yuan, and F. E. N. G. Meng, "Study on the regularity of acupoint selection and compatibility of acupuncture and moxibustion in the treatment of Parkinson's disease based on complex network analysis," Acupuncture Research, vol. 49, no. 12, pp. 1333-1341, 2024.
5. J. Liu, Y. J. Han, T. Y. Zhao, W. R. Li, B. Chen, Y. Guo, and Z. L. Chen, "Rule of acupoint combination in acupuncture treatment of hiccup: a study based on complex network analysis," Zhen ci yan jiu= Acupuncture research, vol. 44, no. 10, pp. 766-771, 2019.
6. Y. Ren, L. Bai, Y. Feng, J. Tian, and K. Li, "Investigation of acupoint specificity by functional connectivity analysis based on graph theory," Neuroscience Letters, vol. 482, no. 2, pp. 95-100, 2010. doi: 10.1016/j.neulet.2010.06.091
7. B. Jiang, Z. Zhang, D. Lin, J. Tang, and B. Luo, "Semi-supervised learning with graph learning-convolutional networks," In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 11313-11320.
8. W. Hamilton, Z. Ying, and J. Leskovec, "Inductive representation learning on large graphs," Advances in neural information processing systems, vol. 30, 2017.
9. M. De Luca, "Artificial intelligence for climate teleconnections: a graph neural network approach with interpretable attention mechanisms (Doctoral dissertation, Politecnico di Torino)," 2025.
10. M. Zitnik, M. Agrawal, and J. Leskovec, "Modeling polypharmacy side effects with graph convolutional networks," Bioinformatics, vol. 34, no. 13, pp. i457-i466, 2018. doi: 10.1101/258814
11. H. Gao, Z. Wang, and S. Ji, "Large-scale learnable graph convolutional networks," In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining, July, 2018, pp. 1416-1424. doi: 10.1145/3219819.3219947
12. Y. Choi, C. Y. I. Chiu, and D. Sontag, "Learning low-dimensional representations of medical concepts," AMIA Summits on Translational Science Proceedings, vol. 2016, p. 41, 2016.
