目的 探究石斛相关保健食品的配方规律，并基于模式生物斑马鱼对含石斛的具有增强免疫力和缓解体力疲劳双功能保健食品的配方进行活性评价。方法 对现有含石斛的保健食品批文中的信息资料进行统计分析，利用社团分析方法，探究石斛相关保健食品中保健功能与中药之间的联系规律；采用关联和复杂网络机器学习算法分析石斛相关保健食品配方中中药与中药之间的关联规则，以期深度挖掘其中的“保健功能-中药-中药”关联性。基于此，设计以石斛为核心原料的具有增强免疫力和缓解体力疲劳双功能保健食品的候选配方；建立2,2′-偶氮二异丁基脒二盐酸盐（AAPH）和甲硝唑诱导的氧化应激斑马鱼模型及机械损伤炎症和神经炎症斑马鱼模型，对其中1种候选配方进行抗炎和抗氧化活性评价。结果 共收集到含石斛的保健食品批文133件，保健功能以增强免疫力和缓解体力疲劳为主。经筛选共得到含石斛且具增强免疫力和缓解体力疲劳双功能的保健食品配方64个，所有保健功能和中药共分为7个社团，归纳为7种治疗原则，得出相应7种中医证型。统计分析发现保健食品配方中高频中药之间具有较强的关联性，并得到了核心中药组合与核心配方，确定10种中药配伍，设计了6个配方。斑马鱼活性实验结果表明其中1种候选配方（石斛、枸杞子和黄芪）对斑马鱼氧化应激损伤和炎症反应均有一定改善作用。结论 机器学习算法指导中药配方合理，为中药类保健食品的配方研发和活性评价提供新的思路和方向。

Objective To explore the formulation regularity of health products containing Shihu (Dendrobii Caulis)and to evaluate enhancing immunity and alleviating physical fatigue functions of health products containing Dendrobii Caulis based on zebrafish model. Methods The information of the approval documents on health products containing Dendrobii Caulis were statistically analyzed. Community analysis method was used to analyze the relation and regulation between health functions and traditional Chinese medicines. association analysis and complex network analysis based on Apiror and multi-scale backbone algorithms were used to analyze the association rules between traditional Chinese medicines in health products formula containing Dendrobii Caulis, in order to deeply explore the relations of “health functions-traditional Chinese medicines-traditional Chinese medicines” of all the traditional Chinese medicines in health products formula with enhancing immunity and alleviating physical fatigue functions. By combining all the analysis results, candidate formulas with enhancing immunity and alleviating physical fatigue functions were designed with Dendrobii Caulisas the main ingredient. Based on 2,2′-azobis (2-methylpropionamidine) dihydrochloride (AAPH)-induced and metronidazole-induced oxidative stress models and inflammation model of zebrafish, anti-inflammatory and anti-oxidant activities of one of the candidate formulas were evaluated. Results A total of 133 health products containing Dendrobii Cauliswere collected in which health function consist largely of enhancing immunity and alleviating physical fatigue. And 64 health products formulations containing Dendrobii Caulis were obtained. All the health functions and traditional Chinese medicines were divided into seven communities, and they were summarized seven therapeutical principles and were classified seven syndrome patterns. It was found that the high-frequency traditional Chinese medicines have a stronger correlation with each other, and the core formulas were obtained based on correlation analysis and complex network analysis. The compatibility of 10 kinds of traditional Chinese medicines was determined and six formulations were designed. The results of zebrafish activity experiment showed that one of the candidate formulas (Dendrobii Caulis, Lycii Fructusand Astragali Radix) had certain ameliorating effects on oxidative stress injury and inflammatory response of zebrafish. Conclusions This study suggests that machine learning algorithm is reasonable to guide the formula of traditional Chinese medicines and it is expected to provide new ideas and methods for the development and activity evaluation of Chinese medicinal health products.

R283.21；R285

北京市科技计划课题（Z201100005420005）；桂林市技术创新引导计划（2020010902）