A pattern or syndrome in response to a multicomponent system is the actual target of herbal medicine treatment. However, it is a substantial challenge to fill the gap between a contributive compound profile in herbal medicine (especially a formula) and its biological features. This study aims to establish a feasible component-mining strategy, which provides a strong prediction of key compounds in support of experimental and clinical observations. Given interdisciplinary scope of life science and mathematical statistics, the relationship between chemical profile and bioactivities is measured by a model termed mathematical prediction bioactivity, in which gray relational analysis, multiple linear/non-linear regression analysis (including t-distributed stochastic neighbor embedding), and radial basis function analysis are involved. R language programming-dependent analysis is adopted with add-on packages, including UniDOE, Factoextra, FactoMineR, Factanal, Rtsne, and Nnet. By using this assessment method in a biological experiment, it is identified that 6-shogaol extracted from Ginger-Coptis formula (a herbal formula) is beneficial for diabetic retinopathy (DR) treatment. The study provides both a novel compound 6-shogalol for DR treatment and a new strategy for mining key contributors in a multicomponent system.

中文翻译：

---

草药中关键作用化合物的鉴定：一种新的数学-生物学评价方法

响应多组分系统的模式或综合征是草药治疗的实际目标。然而，填补草药（尤其是配方）中的有用化合物特征与其生物学特征之间的差距是一项重大挑战。本研究旨在建立一种可行的成分挖掘策略，为支持实验和临床观察的关键化合物提供强有力的预测。鉴于生命科学和数理统计的跨学科范围，化学特征和生物活性之间的关系通过称为数学预测生物活性的模型来衡量，其中灰色关联分析、多元线性/非线性回归分析（包括 t 分布随机邻域嵌入） ，以及径向基函数分析。R 语言编程相关分析采用附加包，包括 UniDOE、Factoextra、FactoMineR、Factanal、Rtsne 和 Nnet。通过在生物实验中使用这种评估方法，确定从姜黄连配方（一种草药配方）中提取的 6-姜酚有利于糖尿病视网膜病变 (DR) 的治疗。该研究既提供了一种用于 DR 治疗的新型化合物 6-没食子醇，也为挖掘多组分系统中的关键贡献者提供了一种新策略。