Potential mechanism prediction of Cold-Damp Plague Formula against COVID-19 via network pharmacology analysis and molecular docking

Background Coronavirus disease 2019 (COVID-19) is a new global public health emergency. The therapeutic benefits of Cold‒Damp Plague Formula (CDPF) against COVID-19, which was used to treat “cold‒dampness stagnation in the lung” in Trial Versions 6 and 7 of the “Diagnosis and Treatment Protocol for COVID-19”, have been demonstrated, but the effective components and their mechanism of action remain unclear. Methods In this study, a network pharmacology approach was employed, including drug-likeness evaluation, oral bioavailability prediction, protein‒protein interaction (PPI) network construction and analysis, Gene Ontology (GO) terms, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway annotation, and virtual docking, to predict the bioactive components, potential targets, and molecular mechanism of CDPF for COVID-19 treatment. Results The active compound of herbs in CDPF and their candidate targets were obtained through database mining, and an herbs—ingredients—targets network was constructed. Subsequently, the candidate targets of the active compounds were compared to those relevant to COVID-19, to identify the potential targets of CDPF for COVID-19 treatment. Subsequently, the PPI network was constructed, which provided a basis for cluster analysis and hub gene screening. The seed targets in the most significant module were selected for further functional annotation. GO enrichment analysis identified four main areas: (1) cellular responses to external stimuli, (2) regulation of blood production and circulation, (3) free radical regulation, (4) immune regulation and anti-inflammatory effects. KEGG pathway analysis also revealed that CDPF could play pharmacological roles against COVID-19 through “multi components‒multi targets‒multi pathways” at the molecular level, mainly involving anti-viral, immune-regulatory, and anti-inflammatory pathways; consequently, a “CDPF—herbs—ingredients—targets—pathways—COVID-19” network was constructed. In hub target analysis, the top hub target IL6, and ACE2, the receptor via which SARS-CoV-2 typically enters host cells, were selected for molecular docking analyses, and revealed good binding activities. Conclusions This study revealed the active ingredients and potential molecular mechanism by which CDPF treatment is effective against COVID-19, and provides a reference basis for the wider application and further mechanistic investigations of CDPF in the fight against COVID-19.


Background
Novel coronavirus pneumonia (NCP) is an infectious disease induced by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which is characterized by a long incubation period, rapid spread, and general population susceptibility. The disease may progress to pneumonia, acute respiratory distress syndrome, acute kidney injury, shock, and multi organ dysfunction [1]. Since January 2020, the virus spread rapidly to most parts of China as well as to other countries worldwide, and the World Health Organization (WHO) declared the disease outbreak as the sixth public health emergency of international concern on January 30, 2020 [2]. NCP was officially named coronavirus disease 2019 (COVID-19) by the WHO on February 11, 2020 [3]. As of May 1, 2020, data released by the WHO revealed 3,175,207 known cases of COVID-19, 224,172 reported deaths, and 211 affected countries or regions [4], these figures are updated daily and are expected to increase further. COVID-19 poses a marked threat to the safety and health of the world's population, and the global impact of the outbreak continues to expand. As no specific drug or vaccine against COVID-19 has been identified to date, creation of a rapid and effective prevention and control program, curbing the spread of disease, and reducing the damage caused by the pandemic are major challenges.
In its long history, Traditional Chinese Medicine (TCM) has accumulated a solid theoretical basis for the treatment of infectious diseases [5], and has recently been implemented against viral pneumonia. TCM have been reported to enhance the immunity of the organism, shorten the duration of treatment, and prevent complications according to numerous studies [6][7][8][9][10]. TCM showed a good treatment effect against severe acute respiratory syndrome (SARS) in 2003. Compared with pure Western medicine-treatment, treatment with pure Chinese herbal medicine could significantly shorten the average absorption time of the lung shadow, without any obvious complications in follow-up, in SARS patients [7]. In the COVID-19 epidemic, China has used a combination of Chinese and Western medicine for both the prevention and treatment of the disease, with notable positive effects. With the strong support of the Chinese government, TCM has been used in the treatment of 91.5% of confirmed cases; inclusion of TCM in COVID-19 treatment not only reduced the death rate of severe/ critical patients by more than 80%, but also resulted in a low rate of relapse in recovered patients; the total effective rate of TCM is more than 90% [11,12]. Further, more and more studies have analyzed mechanisms of action of TCM formulae in treating COVID-19 [13][14][15][16][17]. TCM has been recommended for different stages of COVID-19 treatment since the third trial version of the "Diagnosis and Treatment Protocol for COVID-19" published by National Health Commission of the People's Republic of China, along with Western standard treatment [14,18].
In trial versions 6 and 7 of the "Diagnosis and Treatment Protocol for COVID-19, " the Cold-Damp Plague Formula (CDPF) was recommended for treating "colddampness stagnation in the lung, " as the disease was designated as a "cold-dampness pestilence" based on the clinical features of COVID-19 cases in Wuhan [19]. The formula was devised by academician Xiao-Lin Tong and other experts, particularly for use in suspected and earlystage confirmed COVID-19 cases, and was mainly used in the epicentre of the epidemic, Wuhan, as well as in other large epidemic areas [19][20][21]. CDPF, consisting of 20 herbs (Table 1), is based on several classical prescriptions, including Ma Xing Shi Gan decoction, Ting Li Da Zao Xie Fei decoction, Huo Po Xia Ling decoction, Shen Zhu San, and Da Yuan Yin, and has the TCM functions of ventilating the lungs and expelling pathogenic factors, dispelling toxins and dredging collaterals, dissolving turbidity, strengthening the spleen, and eliminating dampness [19]. More than 50,000 COVID-19 patients, including mild cases, moderate cases, suspected cases, and self-isolating fever cases, have taken CDPF, which showed significant advantages in improving the cure rate, reducing mortality, and promoting physical recovery [12,22].
The effect of TCM is related to the mechanism of the multi-ingredients, multi-targets, and multi-pathways of herbs; however, this multi-faceted mechanism makes it difficult to identify the mechanism of action of CDPF in terms of traditional pharmacological evaluation. On the other hand, network pharmacology is an effective approach to study and clarify the mechanism of drug action, which could explain the mechanism of functional drugs based on a network of public databases or data obtained through early studies. The network pharmacology research strategy is in line with the understanding of disease integrity in TCM [23,24].
Thus, in this study, we utilized network pharmacology and the related technology to explore the main bioactive components of CDPF and predict the effective Keywords: COVID-19, Cold-Damp Plague Formula (CDPF), Network pharmacology, Molecular mechanism, Molecular docking targets of these active components. Consequently, we proposed potential mechanisms underlying the effects of CDPF in the treatment of COVID-19 at a molecular level, which might provide theoretical support for the wider application of CDPF in the fight against COVID-19.

Collection of ingredients of CDPF and screening
The components of herbs used in CDPF were retrieved from the Traditional Chinese Medicine Systems Pharmacology (TCMSP) database (http://tcmsp w.com/tcmsp .php) and Traditional Chinese Medicines Integrated Database (TCMID) (http://www.megab ionet .org/tcmid /) [25], and related references. The candidate ingredients from the TCMSP were retained only if their oral bioavailability (OB) ≥ 30% and if their drug likeness (DL) ≥ 0.18 [26]. Candidate ingredients without potential target information were excluded. The properties of ingredients collected through the TCMID and literature reviews were retrieved from the Swiss ADME database (http://www.swiss adme.ch/). The screening criterion for gastrointestinal GI absorption was set as high, and DL is satisfied with both "yes" at the same time [27]. The structures and SMILES format files of the compounds were obtained from the PubChem (https ://pubch em.ncbi.nlm. nih.gov/) database [28].

Construction of targets related to the identified compounds
To obtain the target of each identified compound, putative targets were predicted from three databases: the TCMSP, similarity ensemble approach (SEA) (http:// sea.bksla b.org) [29], and STITCH (http://stitc h.embl. de/) database [30]. The SMILES format of the identified compounds was uploaded to the STITCH and SEA databases with the "Homo sapiens" setting. Afterward, the target proteins corresponding to the compounds screened from the three target databases were standardized in the UniProt (https ://www.unipr ot.org) database, with the properties set to "reviewed" and "human" [31]. Finally, the targets from the three databases were merged, and the duplicated targets were removed. A Venn Diagram (http://bioin forma tics.psb.ugent .be/ webto ols/Venn/) was used to filter the repeated targets among active components of herbs and to construct the herbs-ingredients-targets network using Cytoscape v.3.2.1 software.

Target dataset for COVID-19
COVID-19-related targets were retrieved from the Human Gene Database (GeneCards, https ://www.genec ards.org/) [32] with the keyword "novel coronavirus pneumonia. " The targets were also sent to the UniProt Database for normalization.

Protein-protein interaction
To clarify the interaction between CDPF-related targets and COVID-19 targets, we screened for overlaps between CDPF-related targets and COVID-19 targetrelated targets. Then, we uploaded the list of overlapping proteins to the Search Tool for the Retrieval of Interacting Genes/Proteins database (STRING) v.11.0 (https :// strin g-db.org/) for PPI analysis [33], with the species limited to "Homo sapiens" and the confidence score cut-off set at 0.4, and the rest of the settings were default. Finally, the final PPIs network was established.

Network construction and analysis
We constructed a PPI network for active ingredients of CDPF and putative COVID-19-related targets of these ingredients based on STRING, and applied Cytoscape v.3.2.1 software to visualize and analyse the interaction network [34]. To study further into the network, the candidates were screened with clustering. Clustering with the Overlapping Neighborhood Expansion (ClusterONE) plug-in of the Cytoscape software was employed to mine the highly interacting gene modules [35], with the minimum size parameter set to 3. The cytoHubba plug-in of the Cytoscape software was used to obtain the top-10 genes, which fetches the shortest path among a group of nodes, based on a mixed character calculation (MCC) score [36].

Gene ontology-enrichment and pathway-enrichment analyses
The candidate genes screened by ClusterONE were further analysed to understand their gene ontology (GO) function, by using the Cytoscape plug-in ClueGO to annotate and visualise the interrelations of terms and functional groups in biological networks. The relevant biological processes (BP), molecular functions (MF), cellular components (CC), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway-enrichment analysis were selected, with a threshold value of p < 0.05 and a kappa score ≥ 0.4.

Molecular docking
Through the above-mentioned cytoHubba analysis, the hub genes for the treatment of COVID-19 were obtained. The top hub gene and other key genes with the more connected active component in herbs were linked by molecular docking. The structural formula (SDF format) of the compounds were downloaded from the PubChem database and converted to PDB format with Open babel v.2.4.1 [37] from the RCSB Protein Data Bank (PDB, https ://www.rcsb.org/) to obtain the crystal structure of the core target [38]. The targets were processed by removing water, adding hydrogen, optimising amino acids, and selecting the magnetic field, and the pdbqt format was saved as a pair acceptor. Atomic charges and assigned atom types were added to compounds in the PDB format, and the pdbqt format was saved as a docked ligand. The active site for molecular docking was determined and size was set. Finally, Autodock Vina v.1.1.2 was run to perform molecular docking [39]. PyMOL v.2.3 software [40], and Molecular Operating Environment v.2.2 (MOE) software [41] were used to visualize the docking results, and based on the binding conformations of the docking results of each compound, the docking results with lower binding energy and better conformation were selected.  [44], and Jiao Mai Ya (JMY, Fructus Hordei Germinatus) [44][45][46] were obtained by literature mining. Next, according to the screening criteria of the ADME, including the OB (or GI absorption) and DL, 193 active ingredients of CDPF were retrieved after duplicated targets were eliminated. Details of component information are provided in the Additional file 1: Table S1.

Target fishing for active CDPF components and construction of the component-target network
The TCMSP, SEA, and STITCH databases were searched for candidate targets of active CDPF ingredients. After fishing for targets, 1172 targets were found for MH, 488 targets were found for KXR, 276 targets were found for QH, 647 targets were found for TLZ, 165 targets were found for GZ, 25 targets were found for XCQ, 397 targets were found for GHX, 211 targets were found for PL, 276 targets were found for CZ, 77 targets were found for FL, 73 targets were found for BZ, 43 targets were found for HP, 371 targets were found for CG, 165 targets were found for SJ, 6 targets were found for DL, 42 targets were found for JBL, 494 targets were found for JMY, 1023 targets were found for JSZ, and 85 targets were found for JSQ. Details of target information are provided in the Additional file 2: Table S2.
The main herbs of CDPF had synergistic effects with each other as exhibited by a Venn diagram. BZ, CG, CZ, FL, GZ, HP, GHX, JMY, JSZ, KXR, MH, PL, QH, SJ, TLZ, and XCQ shared three targets, CG, CZ, GZ, GHX, JMY, JSZ, KXR, MH, PL, QH, and TLZ shared seven targets, BZ, CG, CZ, FL, KXR, MH, PL, QH, SJ, and TLZ shared 12 targets, and CG, GHX, JMY, JSZ, MH, and TLZ shared up to 62 targets, etc. Although there were too many different lists to be captured in a single diagram, the details of the Venn diagram are provided in text form in the Additional file 3: Table S3.
Overall, 949 candidate targets of active CDPF components were identified after removing duplication, and the herbs-ingredients-targets interaction network was consequently constructed using Cytoscape v.3.2.1 software. In order to simplify the network, the nodes above the average degree were selected to build the high connections of "herbs-ingredients-targets" network (Fig. 1). The network containing all the nodes of 949 candidate targets of active CDPF components was shown in Additional file 4: Fig. S1.

Potential COVID-19 targets of CDPF
For disease target identification, 259 target genes that occur in COVID-19 were identified from the GeneCards database (Additional file 5: Table S4). Seventy-one target genes overlapped between CDPF and COVID-19 and were selected as potential targets for further analysis ( Table 2).

PPI network analysis
The STRING database was used to acquire PPI relationships of potential protein targets of CDPF as related to the treatment of COVID-19. By using these targets, the network of PPI relationships was shown to contain 71 nodes (representing active proteins) and 1006 edges (representing the interaction between the active proteins and proteins), with an average node degree of 28.3 and a PPIenrichment p value of < 1.0e−16.
In the PPI network, constructed with the Cytoscape software using parameters such as a minimum required interaction score > 0.4) (Fig. 2a), the most significant module (Density = 0.438, Quality = 0.874, p < 0.001) containing 68 nodes was then recognized by ClusterONE (Fig. 2b).

GO-enrichment analysis
To determine the biological features of the candidate targets screened by ClusterONE, GO analysis was accomplished by the Cytoscape plug-in ClueGO. Based on filter conditions of p < 0.05 and kappa score ≥ 0.4, GO function-enrichment analysis resulted in 466 items, of which BP accounted for 446, CC for 2, and MF for 18 items. BP analysis revealed that the candidate targets were markedly enriched for cellular response to biotic stimulus, inflammatory response, regulation of inflammatory response, reactive oxygen species metabolic process, positive regulation of response to external stimulus, cellular response to external stimulus, cytokine production, regulation of sequence-specific DNA binding transcription factor activity, regulation of chemokine production, reactive oxygen species biosynthetic process, regulation of leukocyte migration, leukocyte homeostasis, B cell proliferation, T cell migration, lymphocyte activation involved in immune response, regulation of nitric oxide biosynthetic process, and response to reactive oxygen species, positive regulation of hemopoiesis, regulation of vasculature development, platelet activation, regulation of blood vessel diameter, positive regulation of cytokine production, regulation of cytokine biosynthetic process, response to extracellular stimulus, response to drug, response to toxic substance, regulation of chemokine biosynthetic process, regulation of production of molecular mediator of immune response, acute inflammatory response, regulation of acute inflammatory response, regulation of immunoglobulin production, receptor-mediated virion attachment to host cell, cellular oxidant detoxification, defense response to protozoan, positive regulation of homeostatic process, and other (Additional file 6: Table S5-1). Here top 20 GO terms of BP were listed by bubble chart according to "term p-value corrected with Bonferroni step down". The CC analysis showed that these candidate genes were enriched in membrane rafts and plasma membrane rafts (Fig. 3b, Additional file 6: Table S5-2). Changes in the MF of the targets were mainly significantly enriched for cytokine receptor binding, phosphatase binding, BH3 domain binding, activating transcription factor binding, chemokine activity, tumour necrosis factor receptor superfamily binding, nuclear hormone receptor binding, R-SMAD binding, and peroxidase activity (Fig. 3c, Additional file 6: Table S5-3).

KEGG pathway analysis
To explore the potential mechanism related to these candidate targets identified by ClusterONE, KEGG pathway analysis was performed. Based on filter conditions of p < 0.05 and kappa score ≥ 0.4, KEGG pathway analysis returned 137 items, including TNF signaling pathway, MAPK signaling pathway, Fc epsilon RI signaling pathway, toxoplasmosis, leishmaniasis, estrogen signaling pathway, natural killer cell mediated cytotoxicity, chagas disease, osteoclast differentiation, longevity regulating pathway, influenza A, sphingolipid signaling pathway, inflammatory bowel disease (IBD), apoptosis, hepatitis B, HTLV-I infection, pertussis, antigen processing and presentation, renin-angiotensin system, and viral myocarditis, AGE-RAGE signaling pathway in diabetic complications, and so on (Fig. 4, Additional file 7: Table S6). These items revealed that the potential pathways affected by CDPF mainly involved improving immunity, anti-inflammatory effects, fighting against viruses and other pathogens.

Discussion
TCM prescriptions are rich in resource-based chemical components, with complex chemical structures, different targets, and diverse action pathways, which work together to exert their unique clinical efficacy. In order to explain the therapeutic basis of CDPF against COVID-19 scientifically, this study revealed the active ingredients and potential molecular mechanism by which CDPF treatment is effective against COVID-19, and provides a reference basis for the wider application and further mechanistic investigations of CDPF in the fight against COVID-19. First, we considered the compound's feasibility for use as a drug, based on the database of TCMSP and TCMID and related references, combining the OB (or GI absorption) and DL values of the compound. Consequently, 193 active compounds of CDPF were assessed and 949 candidate targets of these active compounds were screened through drug target-prediction databases. Venn plots of these candidate targets showed that several herbs shared common targets; for example, the CG, GHX, JMY, JSZ, MH, and TLZ shared up to 62 targets, illustrating the synergy between herbs. We thus constructed an herbsingredients-targets network, which intuitively reflected the characteristics of CDPF, which has multiple components and targets.
We retrieved 259 candidate COVID-19 targets from the GeneCards database. Through target mapping, we found 71 targets in common between CDPF active componenttargets and these disease targets, which were considered potential targets for CDPF in the treatment of COVID-19.
To identify functional connections among these common targets, PPIs were predicted using STRING. We carried out module analysis on a PPI network via the ClusterONE plug-in in Cytoscape; modularization is an important step, as it helps to reduce the noise of the data. By modularization, the most significant cluster (Density = 0.438, Quality = 0.874, p < 0.001) was obtained, which contained 68 seed proteins. We subsequently performed GO and KEGG analyses of these seed proteins.
Using ClueGo enrichment analysis, significant biological processes related to the anti-COVID-19 effects of CDPF were identified. These mainly focused on four areas: (1) Cellular responses to external stimuli, such as cellular response to external stimulus, positive regulation of response to external stimulus, response to extracellular stimulus, cellular response to biotic stimulus, and even receptor-mediated virion attachment to host cell.
(2) Regulation of blood production and circulation, such as regulation of vasculature development, platelet activation, positive regulation of hemopoiesis, and regulation of blood vessel diameter. SARS-CoV-2 may cause blood clotting and trigger the risk of stroke [47][48][49]; the above biological processes may be effective against the occurrence of blood clotting. (3) Free radical regulation, such as reactive oxygen species metabolic process, reactive oxygen species biosynthetic process, regulation of nitric oxide biosynthetic process, and response to reactive oxygen species. Some studies have indicated that high levels of reactive oxygen species and other free radicals are generated after virus invasion, thereby inducing damage to the organism [50][51][52]. These biological functions involve the regulation of free radical production and metabolism. (4) Immune regulation and anti-inflammatory effects, such as cytokine production, regulation of cytokine biosynthetic process, regulation of inflammatory response, regulation of leukocyte migration, regulation of cytokine biosynthetic process, regulation of production of molecular mediator of immune response, regulation of leukocyte migration, leukocyte homeostasis, B cell proliferation, T cell migration, lymphocyte activation involved in immune response, regulation of chemokine production, regulation of acute inflammatory response, regulation of immunoglobulin production, positive regulation of lymphocyte proliferation. In order to fight against a viral invasion, the organism stimulates a series of immune responses. Leukocytes quickly migrate to the site of the virus invasion to surround and phagocytize the virus [53,54]. As the virus continues to replicate, B lymphocytes are activated by antigens, forming plasma cells, and producing immunoglobulins (antibodies), and then antigen-antibody binding participates in the humoral immune response through the blood and body fluids [55,56]. Moreover, a large number of cytokines are released by immune cells to play an anti-viral role; these include ILs, interferons, TNFs, colony stimulating factors, chemokines, and growth factors [57][58][59]. However, while the body produces an immune response to eliminate the virus, the cytokines secreted by the immune cells activate further immune cells, which in turn produce more cytokines, and induce a "cytokine storm. " Eventually, the immune system begins to attack the normal cells of the host [60,61]. This reflects two-way regulation: on the one hand, the body can quickly produce an immune response to play a role in eliminating the virus; on the other hand, it can regulate the production and secretion of cytokines to avoid the occurrence of a cytokine storm, which would cause damage to the organism.
KEGG pathway analysis showed that CDPF mainly regulates disease pathways induced by viruses, such as hepatitis B, influenza A, HTLV-I infection, and viral myocarditis. Additionally, CDPF regulates pathways of diseases caused by external organisms, such as bacteria or worms, including pertussis, leishmaniasis, chagas disease, and toxoplasmosis, etc. Immune and inflammatory pathways, such as antigen processing and presentation, natural killer cell mediated cytotoxicity [62], Fc epsilon RI signaling pathway, associating with the secretion of cytokines, such as TNF-α, contributing to inflammatory responses [63], TNF signaling pathway [64], and MAPK signaling pathway [65], and inflammatory disease-related pathways, such as IBD, are also involved. Sphingolipid signaling pathway utilizes second messenger functions involved in a variety of cellular signalling pathways; sphingolipids appear to be mediators of inflammation and are novel therapeutic targets in inflammatory disease [66]. CDPF also targeted the AGE-RAGE signaling pathway, which is involved in diabetic complications. Diabetes mellitus is a major comorbidity complicating COVID-19, and the association between glycaemic control and prognosis in patients with comorbid COVID-19 and diabetes has been reported [67]. ACE2, a gene involved in the renin-angiotensin system, has been shown to be a therapeutic target in COVID-19, as SARS-CoV-2 uses ACE2 as a receptor to enter host cells [68,69]. In addition, pathways related to apoptosis, longevity, and estrogen signalling are also regulated by CDPF.
Moreover, in the PPI network, the top 10 targets (IL6, TNF, IL10, MAPK8, MAPK3, CXCL8, CASP3, PTGS2, TP53, and MAPK1) were regarded as hub genes that may play important roles in the treatment of COVID-19 by CDPF and are involved in the regulation of immunity and inflammation. For example, TNF has immune regulatory, proinflammatory [70], and anti-viral functions [71], MAPK8 modulates lymphocyte homeostasis [72], IL10 is generally considered to be an anti-inflammatory cytokine [73,74], but production of IL-10 has also been shown to be detrimental during high-dose primary influenza challenge [75]. CXCL8 increases recruitment of principal human neutrophils and is a major inflammatory mediator [76], while PTGS2 is also an inflammatory marker [77]. The most notable factor was IL6, which plays a key role in the cytokine storm [78,79], and is used as a clinical early warning index in the diagnosis and treatment of COVID-19 [80,81]. IL6 plays a central role in the acute inflammatory response, and a long duration of its release can also be used to assess the severity of infection and judge prognosis [82,83]. Dynamic observation of IL-6 levels can assist in understanding the progression of infectious diseases and the response to treatment [84,85].
We then performed docking studies for IL6, using the critical ingredients quercetin and luteolin as ligands, and for ACE2, the receptor by which SARS-CoV-2 typically enters host cells, with its ligands, l-tyrosine and l-phenylalanine. Molecular docking allows assessment of whether it is thermodynamically possible for the ligand and protein to bind. The binding energies of these docking results further helped to refine the targets of CDPF. Finally, the vina scores of the active compounds with the key targets were all negative and less than −6, and their interaction points were revealed, and demonstrated that quercetin and luteolin, and l-tyrosine and l-phenylalanine had good binding activities to IL6 and ACE2, respectively.  6 Network construction of "CDPF-herbs-ingredients-targets-pathways-COVID-19". The "Ellipse" node represents the herb, the "Hexagon" node represents the ingredient, the "Diamond" node represents the target, the "Rectangle" node represents the pathway, and the small white "Triangle" node with big letters represents CDPF and COVID-19