Compatibility law of Chinese medicine for COVID-19
We collected a total of 24 TCM prescriptions including 105 herbs that are highlighted in the guidelines for the treatment of COVID-199. According to the characteristics of the disease progression, we summarized TCM prescriptions that are available at different stages of the disease and their specific herbs in Figure 1. In general, it was found that different TCM prescriptions are recommended at different stages of COVID-19. Even at the same stage, the recommended TCM prescriptions vary according to the specific syndromes of the patients. Among these prescriptions, Qingfei Paidu Decoction is the most recommended prescription, which can be used for mild, moderate and severe patients with COVID-19. Besides, Xuebijing Injection, Reduning Injection, Tanreqing Injection and Xingnaojing Injection were recommended twice, in severe and critical conditions. Besides, it appeared that almost every prescription consists of a variety of different herbs, and the same herb may appear in different prescriptions by certain combination compatibility.
To further explore the prescription rules of TCM treatment for COVID-19, we analyzed 105 kinds of herbs in the recommended TCM prescriptions, and found that the most commonly used herbs were Licorice, Gypsum, Ephedra sinica Stapf, Agastache rugosa, Amygdalus Communis Vas, Baikal Skullcap, and Forsythia suspensa. The top 20 commonly used Chinese herbs were shown in Figure 2A. In order to generate association rules among these herbs, we performed data analysis based on the association rules algorithm in 24 sets of recommended TCM prescriptions. We obtained 32 association rules for herbal combinations from the TCM prescriptions of COVID-19 (Table 1). In general, a big value for the lift (X⇒Y) indicates a stronger association between X and Y21. The lift values of these 32 rules were all greater than 1, this indicated that there were positive interdependence effects on these rules.
Screening of important herbal pairs
We further extracted the combination herb pairs containing two herbs by subset function. There were 16 rules for the combination compatibility of herbal pairs. According to the confidence ranking, {Amygdalus Communis Vas (ACV)}=>{Ephedra sinica Stapf (ESS)}, {Poria cocos}=>{Agastache rugosa}, {ACV}=>{Gypsum}, {Gypsum} => {Licorice}, {ESS} =>{ACV} and {ESS}=>{Gypsum} were the top six important herbal pairs, which were shown in Figure2B. The larger the circle between the two herbs, the higher the confidence level. The darker the color, the higher the lift value. Among them, {ACV}=>{ESS} has the highest confidence degree (1.00) and lift value (3.43), as well as high support degree (0.25). This suggested that the probability of using both herbs at the same time is 25%. Under the premise of using ACV, the probability of using ESS is 100%, which is 3.43 times higher than that of using ESS alone. In addition, further analysis showed that the herbal pair of ACV and ESS was widely used in almost all stages of the disease, including observation period, mild, general, severe and critical conditions. Therefore, the combination of ACV and ESS was selected as the most important herbal pair for the treatment of COVID-19.
The active ingredients of ACV and ESS (AE)
Based on the above analysis, the combination of ACV and ESS was regarded as the most important herbal pair, and its active ingredients and potential targets were further studied. 40 active ingredients from 476 compounds of AE met the requirements of OB ≥ 30% and DL ≥ 0.18. It was found that 5 active components had no corresponding action targets, and the other 35 active ingredients that could be found by literature reviews had a total of 210 targets (Table 2). Among these 35 active ingredients, 21 ingredients were from ESS, 16 ingredients were from ACV, and 2 ingredients ((+)-catechin, stigmasterol) were from both herbs.
“AE-component-target-COVID” network
A total of 261 COVID-19 disease targets were obtained from GeneCards and OMIM database, and 44 crossed targets of AE and COVID-19 were obtained by the R software, as shown in Figure 3A. These crossed targets were regarded as the potential targets of AE against COVID-19. Then, the “AE-component-target-disease” network was constructed by Cytoscape, as depicted in Figure 3C. This network included 73 nodes and 225 edges, with a network density of 0.081 and a network diameter of 4. The key nodes in this network were shown in Table 3. The topological parameters showed that the node-degree distribution obeyed the power-law distribution (Figure 3B). There were 43 ESS targets, 7 ACV targets and 6 overlapped targets (PTGS2, PTGS1, CAT, NOS2, PPARG, and SOD1) from a total of 26 active ingredients of AE. According to the degree value, the most critical ingredients of AE were quercetin, luteolin, kaempferol, naringenin and (+)-catechin, which interacted with 38, 18, 13, 10 and 7 targets of COVID-19, respectively.
PPI network of AE against COVID-19
PPI network has been widely used to identify many different interactions of the protein targets in the context of a complex disease. There were a total of 44 nodes and 324 interaction lines in the STRING PPI network. The PPI enrichment p-value was less than 1.0e-16, demonstrating an obvious protein interaction relationship, which was shown in Figure 4A. Due to the complexity of the original network obtained from the STRING database, we imported the PPI data into Cytoscape to explore the importance of potential targets in the protein networks and the main cluster in this network. The node represents the potential targets, and the larger the node area and the redder the color, the more important the target protein is. As shown in Figure 4B, interleukin 6 (IL-6), mitogen-activated protein kinase (MAPK) 1, MAPK8, interleukin-1β (IL-1β), nuclear factor kappa-light-chain-enhancer of activated B cells (NF-kB) p65 subunit (RELA), C-X-C motif chemokine ligand 8 (CXCL-8), C-C motif chemokine ligand 2 (CCL2) and prostaglandin G/H synthase 2 (PTGS2) were the key target proteins for the treatment of COVID-19 with AE. Among them, IL-6 (degree = 32) was the most critical target in the PPI network, and the main cluster network of IL-6 was shown in Figure 4C.
Enrichment of potential targets of AE
To further explore the underlying mechanisms of AE as a therapy against COVID-19, we performed GO enrichment and KEGG analysis with the 44 potential targets identified by the R software. GO enrichment consists of three parts, biological process (BP), cellular component (CC) and molecular function (MF). There were 1633 GO enrichment terms for BP, and the most enriched terms included responses to lipopolysaccharide (LPS), bacterial molecules, biotic stimulus and oxidative stress, and positive regulation of cytokine production. Besides, a total of 32 CC items were obtained, and the most enriched terms included membrane raft, microdomain and region, caveola, plasma membrane raft, outer membrane, and focal adhesion. There were 78 GO terms for MF enrichment, and the most enriched terms included cytokine receptor binding and activity, receptor ligand activity, chemokine receptor binding, phosphatase binding, MAP kinase activity, and chemokine activity. The top 10 most important GO items for different categories were shown in Figure 5.
From KEGG analysis, we obtained a total of 113 pathways that were mainly divided into several categories, such as human diseases, signal transduction, cell process and immune system. The top 20 significantly enriched KEGG pathways were presented in Figure 6A, including tumor necrosis factor (TNF), Toll-like receptor (TLR), hypoxia-inducible factor-1 (HIF-1), nucleotide-binding oligomerization domain (NOD)-like receptor(NLRs), and several disease-related pathways like Chagas disease and Influenza A. Among these pathways, TNF pathway with relatively lower p-value and FDR value (<0.0001) was regarded as an important pathway of AE against COVID-19. The network of the top 20 pathways and their targets was shown in Figure 6B, where the font size of the label represented the degree value of the node in this network. The detailed information of the gene targets in these 20 pathways was listed in Table 4.
Molecular docking
It is generally believed that the lower the binding energy of ligand and receptor, the more stable the conformation and the greater the possibility of action. The molecular docking results showed that the binding energies of the main active compounds in AE were all less than -5kJ /mol, which indicates that these compounds can well combine with SARS-COV-2 3CL pro and ACE2 to play a role in the treatment of COVID-19. The three compounds with the lowest binding energy to SARS-COV-2 3CL pro were Quercetin, Luteoli, Glabridin. The three compounds with the lowest binding energy to ACE2 were Beta-sitosterol, Stigmasterol, Glabridin. The results are shown in Table 5 and Figure 7.