目的 利用层次分析法(analytic hierarchy process,AHP)-自组织映射(self-organizing map,SOM)聚类-逼近理想解排序(technique for order preference by similarity to solution,TOPSIS)算法和中医传承辅助平台(V2.5),对中医药治疗围绝经期抑郁症组方进行数据挖掘研究,并结合中医理论,挖掘治疗围绝经期抑郁症的新组方。同时证明AHP-SOM聚类-TOPSIS算法可以用于临床疾病处方规律挖掘。方法 收集中国期刊全文数据库(2000-2021年)中中医治疗围绝经期抑郁症的方剂信息,先后使用关联规则算法(Apriori)、AHP、SOM、TOPSIS等关联、决策和聚类机器学习算法,挖掘其中高频原料药味的配方规律,结合传统中医理论得到可能的新组方,并且使用TOPSIS对新方进行综合评价排名。同时,运用中医传承辅助平台(V2.5)进行组方规律分析,得出新方。结果 用药频次分析得到前3位高频药味为柴胡、白芍和甘草,关联规则结果显示高频药味之间产生较强的关联性,排名前3位的药物组合分别为柴胡-白芍、柴胡-甘草、柴胡-茯苓。继而对39个高频药味进行AHP分析和加权后,得到加权值排在前5位的药味为柴胡、白芍、甘草、郁金、半夏,这些药味可以考虑在组方时优先选择。SOM聚类显示所有高频药味可分为7类,其中最优选配方药味与AHP分析结果权重排名前列的药味有极高的重叠。依据传统中医理论中疏肝理气、化痰开窍、活血化瘀等治则进行配伍组合,最终设计了10个可能的配方,并进行TOPSIS分析评价,排名第1的配方为柴胡、酸枣仁、白芍、半夏、郁金、甘草。最后使用中医药传承辅助平台,基于无监督熵层次聚类算法得出2个潜在的核心药组:白芍-麦冬-远志、石菖蒲-柴胡-远志,核心药组再次组合成1个新处方:白芍-麦冬-远志-石菖蒲-柴胡。结论 在中医药基本理论的指导下结合各类机器学习算法,分析治疗围绝经期抑郁症的组方规律,设计获得可能的新组方,为临床治疗围绝经期抑郁症提供新思路。

Objective The analytic hierarchy process (AHP)‐self-organizing map (SOM)‐technique for order preference by similarity to solution (TOPSIS) algorithm and the Traditional Chinese Medicine (TCM) Inheritance Platform System (V2.5) were used to carry out data mining research on TCM treatment of perimenopausal depression, and obtained a new prescription under the guidance of TCM theory. It was proved that the AHP-SOM-TOPSIS algorithm could be used to mine the prescription rules of clinical diseases. Methods The prescription information of perimenopausal depression treated by TCM from China National Knowledge Infrastructure (2000— 2021) were searched. The association rule algorithm (Apriori), AHP, SOM clustering, TOPSIS, and other association, decision making and clustering machine learning algorithms were used to mine the formula rules of high-frequency raw materials, and to obtain possible new formulations combined with traditional Chinese medicine theory. In addition, TOPSIS was used to evaluate and rank the new prescriptions. At the same time, the TCM Inheritance Platform System (V2.5) was used to analyze the rules of prescription formation and obtain the new prescription. Results The analysis of drug frequency showed that the top three high frequency drugs were Chaihu (Bupleuri Radix), Baishao (Paeoniae Radix Alba) and Gancao (Glycyrrhizae Radix et Rhizoma). The top three drug combinations were Bupleuri Radix‐Paeoniae Radix Alba, Bupleuri Radix‐Glycyrrhizae Radix et Rhizoma and Bupleuri Radix‐Fuling (Poria). Then, after the hierarchy analysis and weighting of 39 high-frequency drugs, the top five drugs in term of weighted values were Bupleuri Radix, Paeoniae Radix Alba, Glycyrrhizae Radix et Rhizoma, Yujin (Curcumae Radix), and Banxia (Pinelliae Rhizoma), which should be selected preferentially in the formulation. SOM clustering showed that all high‐frequency drugs could be divided into seven categories, of which the most preferred drugs overlapped with those with the high weighted valules in AHP analysis. According to the treatment principles of TCM, such as soothing the liver and regulating qi, resolving phlegm and resuscitating orifices, invigorating the circulation of blood and resolving stasis, 10 possible formulas were finally designed. The TOPSIS analysis and evaluation showed that the formula ranked first was Bupleuri Radix, Suanzaoren (Ziziphi Spinosae Semen), Bupleuri Radix, Pinelliae Rhizoma, Curcumae Radix, Glycyrrhizae Radix et Rhizoma. Finally, two potential core drug groups were obtained based on the unsupervised entropy hierarchical clustering algorithm using the TCM Inheritance Platform System: Paeoniae Radix Alba-Maidong (Ophiopogonis Radix)‐Yuanzhi (Polygalae Radix) and Shichangpu (Acori Tatarinowii Rhizoma)‐Bupleuri Radix-Polygalae Radix. The above core drug groups were combined into a new prescription: Paeoniae Radix Alba‐Ophiopogonis Radix‐Polygalae Radix‐Acori Tatarinowii Rhizoma‐Bupleuri Radix. Conclusion Under the guidance of the basic theory of TCM, combining with various machine learning algorithms, the prescription rules for the treatment of perimenopausal depression were analyzed, and the possible new prescriptions were designed, as well as provided new ideas for the clinical treatment of perimenopausal depression.

R283.21

河南省重大专项（211110310100）