key: cord-1044223-krevy7gu
authors: Jessen, Walter J.; Gaisenband, Stefan Diaz; Quintanilla, M’Lissa; Lula, Sadiq; McLeroth, Patrick; Sullivan, Adam; Letovsky, Stanley
title: Modeling COVID-19 disease biology to identify drug treatment candidates
date: 2022-04-18
journal: bioRxiv
DOI: 10.1101/2022.04.18.488660
sha: 5f9dc526444d62717ea96cd0e7b9f726ae81d996
doc_id: 1044223
cord_uid: krevy7gu

Coronavirus disease 2019 (COVID-19) is an infectious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Currently, there are a limited number of effective treatments. A variety of drugs that have been approved for other diseases are being tested for the treatment of COVID-19, and thus far only remdesevir, dexamethasone, baricitinib, tofacitinib, tocilizumab, and sarilumab have been recommended by the National Institutes of Health (NIH) COVID-19 Treatment Guidelines Panel for the therapeutic management of hospitalized adults with COVID-19. Using a disease biology modeling approach, we constructed a protein-protein interactome network based on COVID-19- associated genes/proteins described in research literature together with known protein-protein interactions in epithelial cells. Phenotype and disease enrichment analysis of the COVID-19 disease biology model demonstrated strong statistical enrichments consistent with patients’ clinical presentation. The model was used to interrogate host biological response induced by SARS-CoV-2 and identify COVID-19 drug treatment candidates that may inform on drugs currently being evaluated or provide insight into possible targets for potential new therapeutic agents. We focused on cancer drugs as they are often used to control inflammation, inhibit cell division, and modulate the host microenvironment to control the disease. From the top 30 COVID-19 drug candidates, twelve have a role as an antineoplastic agent, seven of which are approved for human use. Altogether, nearly 40% of the drugs identified by our model have been identified by others for COVID-19 clinical trials. Disease biology modeling incorporating disease-associated genes/proteins discussed in the research literature together with known molecular interactions in relevant cell types is a useful method to better understand disease biology and identify potentially effective therapeutic interventions.

As of the 21 st of February, 2022, the coronavirus disease 2019 (COVID-19) has caused over 422, 188, 754 infections and killed over 5,876,766 people worldwide 1 . The SARS-CoV-2 virus that causes COVID-19 was first identified in Wuhan, China, as a novel coronavirus that was related to a cluster of pneumonia cases most commonly associated with respiratory and gastrointestinal symptoms, but can vary in symptoms based on disease severity 2 . While COVID-19 primarily affects the respiratory system, the virus can affect any organ in the body; in critically ill patients, multiple organs are often affected 3 . Most infections will be asymptomatic or mild; however, many people will suffer from moderate to severe infections and a minority will develop acute respiratory distress syndrome (ARDS) requiring critical care support in an intensive care setting.

An epidemiological study across eighteen European countries in 1,420 patients with mild-to-moderate COVID-19 who did not require hospitalization in intensive care units showed that patients with nonsevere disease most commonly suffer from headache (70.3%), loss of smell (70.2%), nasal obstruction (67.8%), cough (63.2%), asthenia (63.3%), myalgia (62.5%), rhinorrhea (60.1%), gustatory dysfunction (54.2%), sore throat (52.9%), and fever (45.4%). However, these symptoms were reported to vary significantly depending on age and sex 4 . Several small studies in early 2020 across local health centers in China revealed that patients admitted to hospital with severe COVID-19 presented most often fever (94-98%), cough (76-79%), dyspnea (55%), and myalgia or fatigue (23-44%) 5, 6 . During the peak of the pandemic, the number of hospitalized patients increased to unsustainable levels across all countries globally, which led to national healthcare systems being overwhelmed and unable to cope with patient volume. Due to the critical situation, the industry had to urgently address unmet medical needs and turned towards drugs that were readily available in the market for the treatment of other conditions. Several large clinical trials have been conducted with marketed drugs, especially in patients that were admitted to the hospital presenting with greater disease severity, and have provided invaluable additional information on the clinical features of COVID-19 and patients' baseline characteristics.

Examples of drugs previously trialed and currently available either under emergency use authorization (EUA) or full approval for the treatment of hospitalized COVID-19 patients that have been repurposed from their use in other indications are baricitinib, a reversible Janus kinase (JAK) inhibitor for treatment of rheumatoid arthritis in adults whose disease was not well controlled by tumor necrosis factor (TNF) inhibitors; dexamethasone, an anti-inflammatory steroid used for many different inflammatory conditions such as allergic disorders and skin conditions; remdesevir, a broad-spectrum antiviral originally developed to treat hepatitis C; tocilizumab, a humanized monoclonal antibody against the interleukin-6 receptor (IL-6R) for the treatment of rheumatoid arthritis and systemic juvenile idiopathic arthritis; tofacitinib, another JAK inhibitor for treatment of rheumatoid arthritis, psoriatic arthritis and ulcerative colitis; and sarilumab, an injectable prescription medicine called an IL-6R blocker for treatment of adults with moderate to severe active rheumatoid arthritis not responding or tolerating treatment with at least one disease-modifying antirheumatic drug 7 . In addition to these drugs, many other compounds have been used in a clinical trial setting to investigate whether they could be used as a potential treatment in COVID-19 infected patients; however, only a minority will be authorized for treatment. Given the global health crisis, more sophisticated ways to develop new drugs against COVID-19 are being explored. 6 We have developed a new methodology to repurpose readily available drugs for a given disease and/or identify candidate biomarkers. This methodology is based on in silico disease biology modelling which aims to recapitulate the molecular environment of diseased target cells to identify candidate biomarkers and/or drugs that can be proposed as options for disease treatment. Others studying SARS-CoV-2 infection and COVID-19 have used protein-protein interaction (PPI) networks to analyze host-virus protein-protein interaction [8] [9] [10] , describe the interactome of the coronavirus S-glycoprotein and host proteins 11 , investigate viral-viral and virus-host PPI networks in peripheral blood mononuclear cells 12 , analyze the immune system PPI network and review potential therapeutic targets for drug repurposing against COVID- 19 13 , and identify potentially repurposable drugs to treat COVID-19 [14] [15] [16] .

Here, we describe the development and analysis of a COVID-19 disease biology model based on genes/proteins significantly associated with the disease discussed in the research literature together with known molecular interactions in epithelial cells to better understand host biological response induced by SARS-CoV-2 and identify effective interventions. We review impacted biological processes/pathways and drug candidates that could be repurposed to treat COVID-19 infection.

To build an interactome network, we first identified relationships between COVID-19 disease and genes/proteins in Medline and PubMed Central using PolySearch 2.0, an online text-mining system 17 .

COVID-19 associations were mined on April 6th, 2021; influenza A H1N1 associations were mined on June 1st, 2021.To ensure consistent identification, our query for COVID-19 included the manually added synonyms coronavirus disease 2019, covid-19 infection, sars-cov-2 infection, and severe acute respiratory syndrome coronavirus 2, and for influenza A H1N1, our query (influenza A H1N1) included the manually added synonyms A/H1N1, swine flu, and H1N1. Default settings were used for both custom filter words and custom negation words. Following both automated mapping and manual review, we identified 14 COVID-19 and 13 influenza A H1N1 genes/proteins, which were used as seeds to construct interactome networks in Ingenuity Pathway Analysis (IPA) (QIAGEN Inc., https://digitalinsights.qiagen.com/products-overview/discovery-insights-portfolio/analysis-andvisualization/qiagen-ipa/) 18 .

Interactome networks were generated using established algorithms and settings, identical for both COVID-19 and influenza A H1N1, based primarily on direct physical molecular interactions. We start by growing up-and down-stream of the seeds, and then look for direct physical interactions between found molecules. We then apply the shortest paths algorithm to the largest single multi-node connected subnetwork and look to add one new molecule (e.g. shortest paths +1) that has been identified experimentally to physically interact with any of the nodes and found molecules.

Influenza A H1N1 is known to infect both epithelial and immune cells 19, 20 . Thus, the only difference in model construction was cell type: COVID-19 used PPI knowledge identified in epithelial cells [21] [22] [23] while influenza A H1N1 used PPI knowledge identified in either epithelial or immune cells. ToppFun, a module of the ToppGene Suite (https://toppgene.cchmc.org) 24 . Correction for multiple testing was performed using the Benjamini-Hochberg (BH) procedure 25 to calculate a false discovery rate (FDR) adjusted p-value (q-value FDR BH). Human phenotypes, diseases, GO:BPs, pathways and drugs were considered significantly enriched if the q-value for the overlap comparison test between input model genes and a given ontology class was less than 0.05. Enrichment analysis typically identifies multiple terms down the same branch of an ontology directed acyclic graph (DAG); nodes further down the DAG provide more information but are less significant and have fewer associated genes than higher nodes. For the top 500 GO:BP terms, we used REVIGO to find a representative subset and group semantically similar terms together 26 . We then selected representative terms from each of the large semantically similar clusters. For diseases, we categorized the top 100 DisGeNET BeFree terms using the Disease Ontology (https://disease-ontology.org/) 27 . For all additional enrichment analyses, to ensure prudent selection of significantly enriched terms and to minimize over-representation, terms were manually reviewed to select those that best balanced context, informative content, and gene enrichment.

To objectively and systematically compare and contrast disease classes instead of individual conditions between COVID-19 and influenza A H1N1 models, the top 100 disease enrichments for each model were mapped from DisGenNET BeFree names to Disease Ontology IDs (DOIDs) 27 . The parent Disease Ontology class(es) was/were used to represent each condition (second level term in the Disease Ontology tree).

Clinical trial information was obtained from Informa Pharma Intelligence's Trialtrove and Clinicaltrials.gov, accessed January 28 th , 2022. To search for studies in Trialtrove, the query (Disease is Infectious Disease: Novel coronavirus (2019-nCoV, COVID-19)) AND [(Tested Drug is X)] was used, which included any alternative names for the drug. Search results were then manually sorted to eliminate any non-relevant results. To search for studies in Clinicaltrials.gov, COVID-19 was entered as the 'condition or disease' and the drug name was entered under 'Intervention/treatment'. Search results were then manually sorted to eliminate any non-relevant results. The results from both searches were then combined and de-duplicated for each drug.

The COVID-19 disease biology model consists of 31 gene/protein nodes (Figure 1 ). After discarding unconnected nodes and small multi-node fragments, of the 14 genes/proteins identified in the literature, six remained in the single large multi-node interactome network. The eight genes/proteins that were discarded included ACE2, SPINK5, FKBP1A, LTB, CEP250, C5AR1, WDHD1, and MIR410; no molecular interactions with these genes/proteins specific to epithelial cells were identified with any other node in the network. Using the same parameters for literature mining and interactome construction, we also built a disease biology model for influenza A H1N1 as a control.

To better understand the biology represented by the COVID-19 model, we performed Gene Ontology (GO) biological process and pathway enrichment analyses. We used REVIGO to remove redundant GO terms for the top 500 significantly enriched biological process terms and grouped semantically similar terms together. For large clusters, we then selected representative terms ( Table 1) . Pathway enrichment analysis incorporated three sources: Gene Map Annotator and Pathway Profiler (GenMAPP), the molecular signatures database (MSigDB C2 BIOCARTA v7.3), and PATHER Pathways ( Table 2) .

To validate and compare the COVID-19 and influenza A H1N1 models, disease enrichment analysis used BeFree text mining data from DisGeNET, a discovery platform containing one of the largest publicly available collections of genes and variants associated to human diseases 28 . Both models met our validation criteria by showing significant enrichment for the modeled disease: the COVID-19 model was significantly enriched with genes/proteins associated with virus diseases (q-value FDR B&H = 2.57E- 13) and coronavirus infections (q-value FDR B&H = 2.95E-03), while the H1N1 model was enriched with genes/proteins associated with virus diseases (q-value FDR B&H = 5.19E-44) and influenza A virus infection (q-value FDR B&H = 2.12E-17). The COVID-19 model was significantly enriched with genes/proteins linked to a number of additional diseases associated with COVID-19 infection ( Table 3) .

The top 100 disease enrichments for each model were then mapped from DisGenNET BeFree names to DOIDs in order to compare/contrast disease classes instead of individual conditions (Figure 2 ).

Indications that could be mapped to multiple high level classes were (e.g., leukemia mapped to both cancer and hematopoietic system disease). We specifically looked at acquired disease indications and their corresponding classes. The COVID-19 model was enriched with more hematopoietic system diseases, including leukopenia and neutropenia, and gastrointestinal system diseases, including diarrhea and cholestasis, than the influenza A H1N1 model. Conversely, influenza A H1N1 model was enriched with more immune system and cardiovascular system diseases than COVID-19.

To provide additional model validation, we also performed human phenotype enrichment analysis ( 

Lastly, to leverage knowledge about interactions between proteins and small molecules, we performed drug enrichment analysis, which incorporated two sources: the Comparative Toxicogenomics Database (CTD) (http://ctdbase.org/) and Stitch (http://stitch.embl.de/) ( Table 5) . We cross-referenced candidate drugs against the PubChem database to identify drug roles shown in in the table. For contextual relevance, we omitted any non-human-use compound (e.g., used as an insecticide, plastics additive, or fluorosurfactant) as well as any compound notated with a legacy record in PubChem (which indicates that the record is no longer maintained by the data contributor and may contain outdated information) or withdrawn from use due to hepatotoxicity.

From the top thirty drug candidates, twelve have a role as an antineoplastic agent. We queried each against the U.S. Food and Drug Administration (FDA) Drugs@FDA database and seven are approved for human use: fulvestrant and ethylnylestradiol target the estrogen receptor; paclitaxel stabilizes tubulin, inhibits the disassembly of microtubules, and blocks mitosis; alitretinoin is a retinoid that binds to and activates nuclear retinoic acid receptors (RAR) and retinoid X receptors (RXR); sirolimus is a macrocyclic lactone that binds to the immunophilin FK Binding Protein-12 (FKBP-12) to generate an immunosuppressive complex that binds to and inhibits the activation of the mammalian Target Of Rapamycin (mTOR), one of the 6 seed genes/proteins identified in the literature and used to build the COVID-19 disease biology model; rifampin inhibits DNA-dependent RNA polymerase in susceptible bacteria; and thalidomide, an immunomodulatory agent that inhibits both the production of tumor necrosis factor alpha (TNF-alpha) in stimulated peripheral monocytes and the activities of interleukins and interferons, inhibits polymorphonuclear chemotaxis and monocyte phagocytosis, and inhibits proangiogenic factors such as vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF), thereby inhibiting angiogenesis. Of the FDA-approved antineoplastic agents, three are planned or in one or more COVID-19 clinical trials as the primary drug being investigated (number of trials in parentheses): alitretinoin (1), sirolimus (7) and thalidomide (7) . The current status of these trials is shown in Table 6 .

There are also clinical trials with non-antineoplastic agents as the primary drug being investigated for the treatment of COVID-19 (number of trials in parentheses): simvastatin (3), ethanol (5), curcumin (17), clotrimazole (18) , ritonavir (272), glutathione (6) , N-acetylcysteine (37) , and hydrogen peroxide (20) . A subset of these agents and trials are being investigated specifically for moderate to severe or critical COVID-19 patients (number of trials in parentheses): simvastatin (3), ethanol (2), curcumin (5), clotrimazole (8), ritonavir (125), glutathione (7), sirolimus (6) , N-acetylcysteine (14) and thalidomide (7); these are shown in Table 7 .

We report here the development and analysis of a COVID-19 disease biology model. The model is based on genes/proteins identified in the research literature together with known molecular interactions in epithelial cells (Figure 1) . We followed a systematic and reproducible strategy to construct the model, and did not manually add genes/proteins, nor include PPIs not present in the knowledgebase. For example, ACE2 is missing from the final model even though it was in the initial list of 14 COVID-19associated genes/proteins identified in the research literature. This is due to a limitation in the IPA knowledgebase; there are no direct physical interactions in epithelial cells up-or down-stream of ACE2 in the knowledgebase. Thus, since it does not connect to any other nodes in our model, it was not included.

Biological process and pathway enrichment analyses provide context to better understand disease biology (Tables 1, 2) . At a high level, these analyses reflect viral infection by way of xenobiotic metabolism, host cellular damage (i.e. response to wounding), defense response and cytokine production, and a change in extracellular-signal-regulated kinase (ERK) signaling, the last step of the Ras/Raf/Mitogen-activated protein kinase/ERK kinase (MEK)/ERK signal transduction pathway. Upon various extracellular stimuli, this regulatory cascade results in ERK-1 and ERK-2 activation, which phosphorylates numerous downstream substrates and leads to the expression of multiple genes essential for diverse cellular functions, such as cell proliferation, differentiation, survival or apoptosis 31,32 . Many of these functions were captured by our model, including apoptosis signaling and negative regulation of apoptotic process, the Ras pathway, leukocyte differentiation, response to oxidative stress, and P53 pathway feedback loops (Tables 1, 2) . Notably, our model captures some of the hematological and neurological manifestations of COVID-19 infection, including blood microparticle formation, coagulation, and cognition.

Transforming growth factor beta (TGF-beta) activates ERK signaling 33 . The TGF-beta signaling pathway is involved in many cellular processes including cell survival, apoptosis and immunity. The TGF-beta signaling pathway was over-represented in our COVID-19 disease biology model ( Table 2) . TGF-beta has been implicated as a key cytokine regulating the ongoing immune activation in severe COVID-19 34 and limits antiviral functions of NK cells 35 .

ERK activation leads to elevated AP-1 activity via c-fos induction 36 . In our COVID-19 disease biology model, the AP-1 transcription factor network is significantly over-represented ( Table 1) The oncostatin M (OSM) signaling pathway was over-represented in our COVID-19 disease biology model ( Table 2) . OSM, a multifunctional cytokine that belongs to the interleukin-6 (IL6) subfamily 42 , was found to be one of the most potent proinflammatory inducers in lung-derived epithelial cells 43 Table 2 ).

In our biological process enrichment analysis, cellular response to oxidative stress, inflammatory response, regulation of immune system process, and positive regulation of receptor signaling pathway via JAK-STAT (Janus kinase/signal transducers and activators of transcription) were all significantly overrepresented in our model ( Table 1) Along with JAK/STAT signaling, the inflammation interleukin signaling pathway was also overrepresented in our COVID-19 disease biology model ( Table 2) . This pathway contains JAK/STAT, as well as the Ras/Raf/MEK/ERK cascade, which couples signals from cell surface receptors to transcription factors to regulate gene expression. The Porcine deltacoronavirus (PDCoV), a newly emerged swine coronavirus that causes acute enteritis in neonatal piglets, was found to activate the Raf/MEK/ERK pathway to promote its replication 54 . Our drug enrichment analysis identified eight drug candidates that play a role as an antioxidant, anti-inflammatory, or immunosuppressive agent -U0126 (1,4-Diamino-2,3dicyano-1,4-bis(o-aminophenylmercapto)butadiene), curcumin, glutathione, sirolimus, chrysin, acetylcysteine, luteolin, and thalidomide -and could potentially be repurposed to treat COVID-19 infection.

In our COVID-19 disease biology model, over-representation of the HIF-1-alpha transcription factor network included HNF4A. The HNF4A protein regulates genes that are important for development and function of beta cells in the pancreas, which produce and secrete insulin. Dysfunction of HNF4A has been associated with diabetes mellitus [55] [56] [57] [58] . HNF4A expression is also detected in epithelia of the pancreas, kidneys, stomach, and intestine, where it exerts functional roles in regulating epithelial junctions and cell proliferation. HNF4A has been found to be a transcriptional sensor of inflammation in the liver and gastrointestinal tract 59 . It also plays a key role in the regulation of angiotensinogen metabolism 60 and has been predicted to regulate intestinal ACE2 expression 61 .

Vitamin D status is associated with risks of influenza and respiratory tract infections, having direct antiviral effects primarily against enveloped viruses, including COVID-19 62 . Gut epithelial vitamin D receptor (VDR) is important in protecting mucosal barrier integrity and regulating the gut innate immunity 63 . In our COVID-19 disease biology model, the VDR pathway was over-represented. A recent systematic review of vitamin D and COVID-19 found that COVID-19 infection, prognosis, and mortality were correlated with vitamin D status 64 .

The Pregnane X receptor (PXR) pathway, also known as the steroid and xenobiotic sensing nuclear receptor (SXR) or nuclear receptor subfamily 1, group I, member 2 (NR1I2), is closely related to VDR and was over-represented in our COVID-19 disease biology model. PXR is a ligand-inducible master transcription factor regulating xenobiotic-and drug-inducible expression of key genes that encode members of the phase I and phase II metabolic enzymes and drug transporters. PXR is expressed mainly in cells of the gastrointestinal tract and liver 65 , and has around 60% homology with VDR in their DNAbinding domains. Thus, PXR can bind to vitamin D-responsive elements in DNA and, as a transcription factor, affect the expression of genes whose expression is normally regulated by vitamin D. PXR activation has been found to suppress the activity of NF-kB, the key regulator of inflammation and immune response [66] [67] [68] [69] . Ligand-activated PXR has been found to interfere with HNF4A signaling 70, 71 , establishing a link between three pathways identified in our pathway enrichment analysis: the HIF-1alpha transcription factor network, VDR pathway, and PXR pathway. Notably, VDR, PXR, and CAR are all members of the NR1I subfamily of nuclear hormone receptors. NR1I subfamily members all share the ability to bind potentially toxic endogenous compounds with low affinity and play major roles in metabolizing xenobiotics and endogenous compounds 72 .

The drugs suggested by the model outputs include seven compounds, three of which are FDA-approved antineoplastic agents ( The semisynthetic antibiotic rifampin is a DNA-dependent RNA polymerase inhibitor and front-line treatment for tuberculosis. Rifampin also binds PXR and is an established PXR agonist commonly used as a prototypical PXR-activating compound 73 ; the PXR pathway was significantly enriched in our pathway analysis ( Table 2 ). In addition, rifampin displays antineoplastic activity in vitro 74, 75 . Rifampin has been identified as a SARS-CoV-2 RNA-dependent RNA polymerase inhibitor using a cell-based SARS-CoV-2 RNA-dependent RNA polymerase activity assay system 76 . A number of in silico screening studies of potent inhibitors against SARS-CoV-2 have also identified rifampin [77] [78] [79] [80] [81] . Rifampicin was also found to have good binding affinity with three cytokines involved in cytokine storm (TNF-alpha, IL-6, and IL-1beta) 80 , suggesting that it may have a poly-pharmacological effect in COVID-19 patients.

Another antineoplastic drug candidate, U0126, inhibits influenza A virus propagation by inducing nuclear retention of the viral ribonucleoprotein complexes (RNPs), impaired function of the nuclear export protein (NEP/NS2), and concomitant inhibition of virus production 82 . U0126 also interferes with the spread of Borna disease virus (BDV) to neighboring cells 83 . Swine acute diarrhea syndrome coronavirus (SADS-CoV) is a newly discovered enteric coronavirus and treatment of cells with U0126 significantly inhibited SADS-CoV infection 84 . U0126 is not currently approved by the FDA for animal or human use.

Various therapeutic effects of curcumin have been reported, including potential chemotherapeutic, antioxidant, antiviral, antibacterial, and anti-inflammatory properties 85 . Nanoencapsulated curcumin treatment in patients with mild and severe disease resulted in a significant reduction in clinical manifestations of COVID-19 (e.g., fever, cough, and dyspnea) and reduced patient mortality rate 86, 87 .

The effectiveness of curcumin-containing nanomicelles as a therapeutic agent for the treatment of COVID-19 is currently being pursued in 17 clinical trials, five of which are for moderate to severe or critical COVID-19 ( Table 7) .

Candidate non-antineoplastic agents enriched in the model may also be useful for the treatment of COVID-19. For example, simvastatin, widely used as a hydroxymethylglutaryl (HMG)-CoA reductase inhibitor to lower blood cholesterol, was associated with a lower risk of developing severe COVID-19 and a faster time to recovery among hospitalized patients without severe disease if used during the 30 days prior to hospital admission 88 . In a retrospective observational study of hospitalized COVID-19 patients taking simvastatin completed in 2020 (ClinicalTrials.gov identifier NCT04407273) there was a significantly lower mortality rate in patients on statin therapy than the non-statin group, and the mortality rate was even lower in patients who maintained their statin treatments during hospitalization 89 . Three trials of simvastatin for COVID-19 treatment, all for moderate to severe or critical COVID-19, are currently being pursued ( Table 7) .

Acetylcysteine, also known as N-acetylcysteine (NAC), is a precursor of the antioxidant glutathione (another drug candidate in Table 5 ) and has been in clinical use for more than 50 years as a mucolytic drug. NAC treatment in patients with influenza was found to significantly decrease the frequency of influenza-like episodes, as well as the severity and duration of most symptoms 90 . Subsequent research found that NAC inhibited human influenza virus (H5N1, Vietnam/VN1203 strain) replication in human lung epithelial cells in a dose-dependent manner and reduced the production of pro-inflammatory cytokines, thus reducing chemotactic migration of monocytes 91 . Thirty-seven clinical trials of NAC, fourteen of which are for moderate to severe or critical COVID-19 (Table 7) , are currently being pursued.

There are several limitations to our COVID-19 disease biology model. The discovery of relevant molecular targets and protein interactions is dependent on existing knowledge; the model cannot make use of existing biological knowledge if it isn't cataloged in the knowledgebase. On that note, scientific rigor and data quality of published research can have a direct impact on the model; we rely on the ensemble of molecules and interactions to outweigh any potentially poor quality data inputs. As COVID-19 can impact multiple organs/tissue types, our model may fail to identify important PPIs because they cannot be recapitulated in epithelial cells. Lastly, our model is a static representation of interactions, neglecting the temporal and spatial organization of protein dynamics.

In conclusion, our study not only presents a systematic examination of host targets and cellular networks perturbed in COVID-19 patients, revealing insights into the mechanisms by which SARS-CoV-2 can trigger a cytokine storm, but also provides a list of candidate drugs, representing a powerful resource in the pursuit of therapeutic interventions. Table 6 . COVID-19 clinical trial status with model-derived FDA-approved antineoplastic agents. 

World Health Organization Coronavirus (COVID-19) Dashboard

Predictors of COVID-19 severity: A literature review

Effect of COVID-19 on the Organs. Cureus

Clinical and epidemiological characteristics of 1420 European patients with mild-to-moderate coronavirus disease 2019

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

COVID-19 Treatment Guidelines -Therapeutic Management of Hospitalized Adults With COVID-19

Interactomes of SARS-CoV-2 and human coronaviruses reveal host factors potentially affecting pathogenesis

A SARS-CoV-2 protein interaction map reveals targets for drug repurposing

Exploring the SARS-CoV-2 virus-host-drug interactome for drug repurposing

COVID-19: viral-host interactome analyzed by network based-approach model to study pathogenesis of SARS-CoV-2 infection

Virus-Host Interactome and Proteomic Survey Reveal Potential Virulence Factors Influencing SARS-CoV-2 Pathogenesis

Single-Cell RNA Sequencing of Human Tissues, and Artificial Neural Networks Reveal Potential Therapeutic Targets for Drug Repurposing Against COVID-19

In silico drug repurposing in COVID-19: A network-based analysis

Designing a Network Proximity-Based Drug Repurposing Strategy for COVID-19

Repurposing novel therapeutic candidate drugs for coronavirus disease-19 based on protein-protein interaction network analysis

PolySearch2: a significantly improved text-mining system for discovering associations between human diseases, genes, drugs, metabolites, toxins and more

Causal analysis approaches in Ingenuity Pathway Analysis

Respiratory epithelial cells in innate immunity to influenza virus infection

Influenza virus directly infects human natural killer cells and induces cell apoptosis

SARS-CoV-2 Receptor ACE2 Is an Interferon-Stimulated Gene in Human Airway Epithelial Cells and Is Detected in Specific Cell Subsets across Tissues

The protein expression profile of ACE2 in human tissues

Tissue distribution of ACE2 protein, the functional receptor for SARS coronavirus. A first step in understanding SARS pathogenesis

ToppGene Suite for gene list enrichment analysis and candidate gene prioritization

COVID-19 induces a TGF-β-dominated chronic immune response that does not target itself

Untimely TGFβ responses in COVID-19 limit antiviral functions of NK cells

The regulation of AP-1 activity by mitogen-activated protein kinases

Mining of high throughput screening database reveals AP-1 and autophagy pathways as potential targets for COVID-19 therapeutics

AP-1 subunits: quarrel and harmony among siblings

Involvement of Autophagy in Coronavirus Replication

The ER stress sensor IRE1 and MAP kinase ERK modulate autophagy induction in cells infected with coronavirus infectious bronchitis virus

Human Coronavirus: Host-Pathogen Interaction

Oncostatin M, a multifunctional cytokine

Stimulatory effect of inflammatory cytokines on alpha 1

COVID-19 as a mediator of interferon deficiency and hyperinflammation: Rationale for the use of JAK1/2 inhibitors in combination with interferon

Signaling to Control Cytokine Release Syndrome in COVID-19

JAK-STAT Pathway Inhibition and their Implications in COVID-19 Therapy

Porcine deltacoronavirus activates the Raf/MEK/ERK pathway to promote its replication

Mutations in the human genes encoding the transcription factors of the hepatocyte nuclear factor (HNF)1 and HNF4 families: functional and pathological consequences

The role of HNF4A variants in the risk of type 2 diabetes

Mechanisms linking obesity to insulin resistance and type 2 diabetes

HNF4A genetic variants: role in diabetes. Curr Opin Clin Nutr Metab Care

Hepatocyte nuclear factor 4-alpha involvement in liver and intestinal inflammatory networks

Regulated Expression of Human Angiotensinogen Gene by Hepatocyte Nuclear Factor 4 and Chicken Ovalbumin Upstream Promoter-Transcription Factor

Bioinformatic characterization of angiotensin-converting enzyme 2, the entry receptor for SARS-CoV-2

Vitamin D and the anti-viral state

Vitamin D/Vitamin D Receptor Signaling Is Required for Normal Development and Function of Group 3 Innate Lymphoid Cells in the Gut. iScience

Effects of Vitamin D on COVID-19 Infection and Prognosis: A Systematic Review

Identification of a human nuclear receptor defines a new signaling pathway for CYP3A induction

Control of steroid, heme, and carcinogen metabolism by nuclear pregnane X receptor and constitutive androstane receptor

The steroid and xenobiotic receptor (SXR), beyond xenobiotic metabolism

Targeting xenobiotic receptors PXR and CAR for metabolic diseases

Role of CAR and PXR in xenobiotic sensing and metabolism

Ligand-activated pregnane X receptor interferes with HNF-4 signaling by targeting a common coactivator PGC-1alpha. Functional implications in hepatic cholesterol and glucose metabolism

Mechanism of rifampicin and pregnane X receptor inhibition of human cholesterol 7 alpha-hydroxylase gene transcription

Evolution and function of the NR1I nuclear hormone receptor subfamily (VDR, PXR, and CAR) with respect to metabolism of xenobiotics and endogenous compounds

Coordinate regulation of human drug-metabolizing enzymes, and conjugate transporters by the Ah receptor, pregnane X receptor and constitutive androstane receptor

Inhibition of cancer progression by rifampicin: involvement of antiangiogenic and anti-tumor effects

Rifampicin as an oral angiogenesis inhibitor targeting hepatic cancers

SARS-CoV-2 RdRp Inhibitors Selected from a Cell-Based SARS-CoV-2 RdRp Activity Assay System

In silico screening of potent inhibitors against COVID-19 key targets from a library of FDA-approved drugs

Potential inhibitors of SARS-cov-2 RNA dependent RNA polymerase protein: molecular docking, molecular dynamics simulations and MM-PBSA analyses

Prediction of potential inhibitors for RNA-dependent RNA polymerase of SARS-CoV-2 using comprehensive drug repurposing and molecular docking approach

Rifampicin and Letermovir as potential repurposed drug candidate for COVID-19 treatment: insights from an in-silico study

Potential RNA-dependent RNA polymerase inhibitors as prospective therapeutics against SARS-CoV-2

Influenza virus propagation is impaired by inhibition of the Raf/MEK/ERK signalling cascade

MEK-specific inhibitor U0126 blocks spread of Borna disease virus in cultured cells

Swine acute diarrhea syndrome coronavirus replication is reduced by inhibition of the extracellular signal-regulated kinase (ERK) signaling pathway

The dual role of curcumin and ferulic acid in counteracting chemoresistance and cisplatin-induced ototoxicity

Immunomodulatory effects of nanocurcumin on Th17 cell responses in mild and severe COVID-19 patients

Nano-curcumin therapy, a promising method in modulating inflammatory cytokines in COVID-19 patients

Relation of Statin Use Prior to Admission to Severity and Recovery Among COVID-19 Inpatients

EFFECT oF STATIN THERAPY oN SARS-CoV-2 INFECTION-RELATED. Eur Hear journal Cardiovasc Pharmacother

Attenuation of influenza-like symptomatology and improvement of cell-mediated immunity with long-term N-acetylcysteine treatment

N-acetyl-l-cysteine (NAC) inhibits virus replication and expression of pro-inflammatory molecules in A549 cells infected with highly pathogenic H5N1 influenza A virus

Anthrax lethal factor endopeptidase inhibitor; prodrug; hydroxymethylglutaryl-CoA reductase inhibitor

D001647 Bile acids and salts 439520 Digestion and absorption of lipids and fat-soluble vitamins 8

Antineoplastic agent; estrogen receptor antagonist

D000431 Ethanol 702 Antiseptic drug; polar solvent; neurotoxin; central nervous system depressant; teratogenic agent; NMDA receptor antagonist; protein kinase C agonist; disinfectant; human metabolite; Saccharomyces cerevisiae metabolite

Metabolite; anti-inflammatory agent; antineoplastic agent; hepatoprotective agent; flavouring agent; biological pigment; nutraceutical; antifungal agent; dye; lipoxygenase inhibitor; ligand; radical scavenger

histone deacetylase inhibitor; immunomodulator; iron chelator; neuroprotective agent; food colouring; aldehyde reductase inhibitor; shikimate dehydrogenase inhibitor

NAD(P)H dehydrogenase

No grants or financial support were used for this study.

All authors are employed by Laboratory Corporation of America Holdings (Labcorp), some of whom own Labcorp stock.

 -06  8  377  JUN, C3, TNF, KLF4, EGFR, F2R, IL6, CD36   GO:0000165 MAPK cascade  3.17E-06  11  1020  KRAS, JUN, C3, TNF, WNT4, HSF1, KLF4, EGFR, F2R, IL6,  CD36  GO:0006954 inflammatory response  6.39E-06  10  862  C3, SERPINA1, TNF, KLF4, EGFR, F2R, FOS, IL6, CD36, SMAD3   GO:0002682  regulation of immune system  process  1.09E-05  13  1747  GATA2, KRAS, JUN, C3, TNF, VTN, CREB1, MTOR, FOS, IL6,  CD36, TP53, Alitretinoin  0  1  0  0  0  0  1   Sirolimus  3  1  0  3  0  0  7   Thalidomide  1  3  2  0  1  0  7