抗结核药导致的肝损伤是结核患者无法规律用药、治疗失败的重要因素。中医药疗法对于干预抗结核药致肝损伤具有独特的优势，临床报道众多，治疗策略灵活而丰富，但缺乏系统整理与挖掘。复杂系统熵聚类算法善于处理挖掘非线性、复杂性的中医数据。本项目拟对近年来国内外中药干预抗结核药导致肝损伤的临床报道文献进行系统整理与评价，构建保肝方药文献数据库。采用复杂系统熵聚类等多种无监督数据挖掘方法对数据进行分析，再结合多名专家评定意见，初步获得干预抗结核药致肝损伤的核心方药。同时构建斑马鱼肝损伤模型，通过检验肝损生化指标、肝脏形态、肝损特异性蛋白等，对数据挖掘出的核心方药药效进行高效初筛，在初筛的基础上利用大鼠肝损模型进行进一步药效验证，最终确定核心方药，总结中医药疗法干预抗结核药致肝损伤的中医辨治规律，为临床用药提供参考。将数据挖掘筛选与实验验证筛选相结合，探索构建药物筛选新模式。

DILI caused by anti-tuberculosis drugs is an important factor for tuberculosis patients to fail in regular drug use and treatment. TCM therapy has unique advantages in intervening liver injury caused by anti-tuberculosis drugs. There are many clinical reports, flexible and abundant treatment strategies, but lack of systematic collation and excavation. Entropy clustering algorithm of complex system is good at processing and mining non-linear and complex data of TCM. This project intends to systematically collate and evaluate the clinical reports on DILI caused by anti-tuberculosis drug intervention in recent years, and construct the literature database of Baogan formulas and herbs. The data were analyzed by complex system entropy clustering and other unsupervised data mining methods, and then the core formulas and herbs for interfering with DILI caused by anti-tuberculosis drugs were obtained preliminarily by combining with the evaluation of experts. At the same time, the zebrafish liver injury model was constructed. By examining the biochemical indicators of liver damage, liver morphology, liver damage specific proteins and so on, the drug efficacy of the core formulas and herbs excavated from the data was screened efficiently. On the basis of the preliminary screening, the rat liver injury model was used to further validate the drug efficacy. Finally, the core formulas and herbs were determined, and the differentiation and treatment of TCM for interfering with DILI caused by anti-tuberculosis drugs were summarized. Regularity can provide reference for clinical medication. Combining data mining screening with experimental validation screening, a new drug screening model was explored.