2. Materials and Methods

2.1. Targets Mining of Main Active Components of QSYQ
The main active components of QSYQ were used to study the mechanism of QSYQ. By literature retrieval from PubMed and CNKI database, the main active components of QSYQ were obtained based on the principles that components are the main efficacy components, have rich content, and can be absorbed into the blood. The information of the main active components of QSYQ is shown in Table 1.
The targets' information of main active components of QSYQ was obtained from two parts: pharmacophore virtual screening and the component-protein interaction database including ChEMBL (https://www.ebi.ac.uk/chembl/#) [11] and STITCH 3.1 (http://stitch.embl.de/) [12]. The 27 pharmacophore models which were applied to virtual screen were constructed by our laboratory team [13, 14]. ChEMBL is a manually curated chemical database which contains compound bioactivity data against drug targets. STITCH is a database in which every interaction has a confidence score, and the interactions with a confidence score > 0.7 were selected.

2.2. Network Construction of Single Herb and Formula
The PPIs information of targets was obtained from the online updated database of String 9.1 (http://string-db.org/) which has a confidence score for every protein interaction [15]. PPIs with a confidence score > 0.7 were applied to construct PIN using Cytoscape which is one of the most popular open-source software tools for the visual exploration of biomedical networks composed of protein, gene, and other types of interactions [16]. Every single herb network is formed only by PPIs involving proteins of this herb, and the formula network is formed only by PPIs involving proteins of this formula.

2.3. Network Analysis
The analysis of topological properties based on topological parameters has become very popular for gaining insight into the organization and structure of the resultant large complex networks [17–19]. Therefore, the topological parameters such as degree distribution, average shortest path, and clustering coefficient were analyzed by Network Analyzer [20] in Cytoscape. Properties of scale-free, small word, and modularity of the QSYQ's PIN were also investigated.
Functional modules of the network were explored by the MCL [21] which simulates a flow on the graph by calculating successive powers of the associated adjacency matrix and the value of the inflation parameter strongly influences the number of clusters. Compared to the other algorithms, for example, RNSC [22], MCODE [23], and SPC [24], the MCL is superior with highlighting the robustness to graph alterations [25]. Based on the identified modules, GO enrichment analysis was utilized to predict possible biological roles of the modules by evaluating the involved biological processes, using the BinGO [26] plugin for Cytoscape.