SZBC-AI4TCM: a comprehensive web-based computing platform for traditional Chinese medicine research and development

Jidong Lang, Kaimin Guo, Jinna Yang, Pengcheng Yang, Yu Wei, Jin‐Ping Han, Shuang Zhao, Zhihong Liu, He Yi, Xin Yan, B Chen, Chengshan Wang, Jian Xu, Jiawei Ge, Wen Zhang, Xuezhong Zhou, Jiansong Fang, Jing Su, Kaijing Yan, Yunhui Hu, Wenjia WangFrontiers in Pharmacology 2025 · DOI

TL;DR

Traditional Chinese Medicine (TCM) research faces challenges due to complex data and the need for interdisciplinary expertise. This paper introduces SZBC-AI4TCM, a comprehensive web-based platform that integrates AI and bioinformatics tools to streamline TCM research and development, making it more efficient and accessible.

What Problem Does This Paper Address?

Traditional Chinese Medicine (TCM) is a rich system with a long history, but its complexity poses significant hurdles for modern scientific research. Unlike Western medicine's focus on single targets, TCM often involves multi-component, multi-target approaches, leading to intricate data that is difficult to analyze using conventional methods. Researchers often need expertise in multiple programming languages, AI, and bioinformatics, creating a barrier to entry. Furthermore, existing computational tools can be difficult to deploy, lack user-friendliness, and require significant computational resources. This gap hinders the modernization and efficient development of TCM-based therapies.

Methods

The SZBC-AI4TCM platform is built upon the WeMol computational framework, which provides a robust architecture for managing analytical modules, workflows, and computational jobs. Key features include:

  • Integrated AI and Bioinformatics Tools: The platform incorporates a wide array of modules for tasks such as TCM formulation analysis, mechanism elucidation, drug screening, protein analysis, small-molecule analysis, and network analysis.
  • User-Friendly Web Interface: A visual interface allows researchers, even those without programming backgrounds, to conduct complex analyses interactively.
  • Hardware-Software Acceleration: The platform leverages powerful hardware (e.g., NVIDIA GeForce RTX 2080 Ti GPU) to expedite data processing.
  • "ShuZhiBenCao" (Digital Herbal) Concept: This guiding principle integrates AI to accelerate TCM research.
  • Max Knowledge Base (MaxKB): An open-source QA system powered by LLMs and retrieval-augmented generation is integrated to assist users in understanding modules and resolving queries.
  • Specific Analytical Modules:
    • Association Rules Analysis: Identifies synergistic relationships between botanical drugs.
  • Network Pharmacology Analysis: Maps herbal metabolites to targets and analyzes their interactions with disease genes.
  • Molecular Docking: Predicts binding modes and energies between metabolites and receptor proteins.
  • Drug Screening: Utilizes gene expression data and databases like LINCS L1000 and ITCM to identify potential therapeutic compounds.

Datasets

The platform integrates several databases and utilizes publicly available datasets for its analyses:

  • Syndrome-Prescription Database: Contains 3,716 TCM formulations.
  • HIT2 Database: Used for herbal metabolite extraction.
  • DisGeNET, eDGAR, GWAS, Pharos, Open Targets Platform: Used for disease-gene association retrieval.
  • STRING Database: For protein-protein interaction network analysis.
  • Gene Expression Omnibus (GEO): Specifically, the GSE159699 dataset for RNA-seq raw count data related to Alzheimer's disease.
  • PubChem Database: For retrieving SDF files of ligands.
  • LINCS L1000 and ITCM Databases: Used for drug screening.

The paper does not introduce new benchmark datasets but rather leverages existing ones for its case studies.

Key Insights and Ideas

  • Bridging the Expertise Gap: SZBC-AI4TCM democratizes TCM research by providing a user-friendly interface that abstracts away complex programming and bioinformatics requirements, empowering wet-lab researchers.
  • Holistic TCM Analysis: The platform's integrated approach, from formulation analysis to drug screening and mechanism elucidation, allows for a more comprehensive understanding of complex TCM systems.
  • Accelerating Discovery: By automating and streamlining complex analyses, the platform significantly reduces the time, labor, and cost associated with TCM research and drug development.
  • Uncovering Synergistic Interactions: The association rules analysis effectively highlights synergistic relationships between botanical drugs, offering crucial insights into TCM formulation principles.
  • Validating TCM Efficacy: The platform's ability to integrate network pharmacology and molecular docking provides a robust framework for validating the multi-target mechanisms of TCM formulations.

Key Results

  • Alzheimer's Disease Case Study: The platform was successfully applied to Alzheimer's disease (AD) research, identifying 399 relevant formulations.
  • Botanical Drug Association Rules: 61 statistically significant botanical drug association rules were identified, with rules like "{Ren Shen, Fu Ling, Fang Feng => Rou Gui}" showing high confidence and lift values, indicating strong synergistic potential.
  • Shuyu Wan Analysis: A representative AD formulation, "Shuyu Wan," was analyzed, revealing 60 overlapping genes between its targets and AD-related genes, enriched in key pathways like PI3K-Akt and MAPK signaling.
  • Network Proximity: Analysis of Shuyu Wan showed that 11 out of 14 botanical drugs exhibited strong AD-targeting efficacy (Zscore < -3).
  • Molecular Docking: Docking simulations demonstrated strong binding capacities for metabolites from Shuyu Wan with ACHE and PPARG receptors, with many complexes showing energies < -5 kcal/mol.
  • Drug Screening: The platform identified potential therapeutic compounds for AD, with six out of the top 10 LINCS small molecules previously reported to have anti-AD effects.

Limitations and Future Work

  • Data Standardization: The paper acknowledges that TCM names can vary significantly, leading to potential issues with information retrieval. They are developing a comprehensive TCM synonym dictionary and plan to use LLM-based RAG for name normalization.
  • Multilingual Support: While some modules support multilingual inputs, others (like the Herbs Database) currently only support Chinese, which can impact user experience and analytical accuracy.
  • Data Updates: Databases are not synchronized in real-time due to resource and cost implications. Updates are planned quarterly or more frequently based on user demand.
  • Integration and Validation: The platform currently performs analyses independently. Future work will focus on deeper integration and comprehensive evaluation of results, emphasizing the need for experimental validation (in vitro, animal models, clinical trials).
  • Computational Resource Demands: The exponential growth in computational needs with increasing module deployment necessitates systematic resource scaling.
  • Learning Curve: Advanced features may present a non-trivial learning curve for some users, requiring targeted training.
  • LLM Integration: The intelligent Q&A functionality is under development and could benefit from integration with additional LLMs.
  • Data Completeness: The quality and completeness of foundational data require continuous updates.

Summary updated

Abstract

SZBC-AI4TCM not only provides robust computational support for TCM research but also significantly enhances efficiency and reduces costs. It offers novel approaches for studying complex TCM systems, thereby advancing the modernization of TCM. As interdisciplinary collaboration and cloud computing continue to evolve, SZBC-AI4TCM is poised to play a strong role in TCM research and foster its growth in addition to contributing to global health. SZBC-AI4TCM is publicly for access at https://ai.tasly.com/ui/\#/frontend/login.
BibTeX
@inproceedings{lang2025szbc,
  author = {Jidong Lang and Kaimin Guo and Jinna Yang and Pengcheng Yang and Yu Wei and Jin‐Ping Han and Shuang Zhao and Zhihong Liu and He Yi and Xin Yan and B Chen and Chengshan Wang and Jian Xu and Jiawei Ge and Wen Zhang and Xuezhong Zhou and Jiansong Fang and Jing Su and Kaijing Yan and Yunhui Hu and Wenjia Wang},
  title = {{SZBC-AI4TCM: a comprehensive web-based computing platform for traditional Chinese medicine research and development}},
  booktitle = {Frontiers in Pharmacology},
  year = {2025},
  doi = {10.3389/fphar.2025.1698202},
}
Figure