基于重叠社区的“方-药”网络经方配伍规律分析

Combination rule analysis of classic formulae in “formulae-herb” network based on overlapping community

针对传统数据挖掘技术难以揭示海量方剂配伍复杂性的问题,借鉴复杂网络思想构建"方-药"网络并基于重叠社区分析其组方配伍规律。首先将方剂内包含的药物抽象为节点、药间关系抽象为边、药物共现频次表示为边的权重,构建"方-药"网络模型,引入"药物影响力"的概念识别网络中的重要药物节点;其次提出一种基于重叠社区的"方-药"网络划分方法进行社区发现;最后分析划分后社区中蕴含的组方配伍规律。在基于《伤寒论》112首方剂构建的模型中,药物影响力大的药物符合张仲景"寒温并用、辛甘苦同行"的用药原则,结合FCM(fuzzyC-means)算法可发现隐藏在《伤寒论》整体模型中的常用药组;在基于《伤寒论》少阴病证构建的模型中,药物影响力大的药物符合少阴病证"温经扶阳、利水渗湿"的用药特点,并得到了分类明显的药物配伍社团。"方-药"网络模型中蕴含着高质量的重叠社区结构,分析社区中蕴含的内在规律可为方剂组方配伍及用药规律的探究提供新的依据。

The compatibility of Chinese materia medica(CMM) studies the relations between herbs and their interacting effects. With complex medications for individuals and massive CMM data, it is difficult for traditional data mining methods to reveal the inherent rules in compatibility analysis. This paper with the theory of complex network constructs a "formulae-herb" network to analyze herb combinations based on the overlapping communities for internal rules in CMM classic formulae. Based on the complex network, the formulae can be abstracted to a graph and represented by nodes and edges, herbs as nodes, and the relationships of herbs as edges. The edges were weighted according to the frequency of co-occurrence of two herbs. "Formulae-herb" network was then constructed based on all formulae in the study. A definition "herb influence" was introduced into the network to identify important herb nodes, with which a method of "formulae-herb" network division was introduced based on overlapping communities to discover the structure of the herb relations. Finally, the inherent laws in the community were discussed. A "formulae-herb" network was constructed based on 112 formulae from Treatise on Cold Damage. Herbs with high influence are in line with Zhang Zhong-Jing’s medication principle of "combining cold and warm, use of pungent, sweet and bitter together". This paper also incorporates FCM(fuzzy C-means) algorithm to discover the commonly used herb groups hidden in the medication. A model of treating Shaoyin syndrom was constructed as an application. Herbs with high influence were in line with the treating method of "warming meridian and reinforcing yang, promoting urination and draining dampness". The result showed that multiple herb groups were clearly classified. "Formulae-herb" network model showed high quality overlapping community structures, the analysis of inherent laws in the community can provide basis for the exploration of compatibility rules of herbs and internal rules in formulae regularities.