Network-based drug repurposing for novel coronavirus 2019-nCoV/SARS-CoV-2

Abstract

Human coronaviruses (HCoVs), including severe acute respiratory syndrome coronavirus (SARS-CoV) and 2019 novel coronavirus (2019-nCoV, also known as SARS-CoV-2), lead global epidemics with high morbidity and mortality. However, there are currently no effective drugs targeting 2019-nCoV/SARS-CoV-2. Drug repurposing, representing as an effective drug discovery strategy from existing drugs, could shorten the time and reduce the cost compared to de novo drug discovery. In this study, we present an integrative, antiviral drug repurposing methodology implementing a systems pharmacology-based network medicine platform, quantifying the interplay between the HCoV-host interactome and drug targets in the human protein-protein interaction network. Phylogenetic analyses of 15 HCoV whole genomes reveal that 2019-nCoV/SARS-CoV-2 shares the highest nucleotide sequence identity with SARS-CoV (79.7%). Specifically, the envelope and nucleocapsid proteins of 2019-nCoV/SARS-CoV-2 are two evolutionarily conserved regions, having the sequence identities of 96% and 89.6%, respectively, compared to SARS-CoV. Using network proximity analyses of drug targets and HCoV-host interactions in the human interactome, we prioritize 16 potential anti-HCoV repurposable drugs (e.g., melatonin, mercaptopurine, and sirolimus) that are further validated by enrichment analyses of drug-gene signatures and HCoV-induced transcriptomics data in human cell lines. We further identify three potential drug combinations (e.g., sirolimus plus dactinomycin, mercaptopurine plus melatonin, and toremifene plus emodin) captured by the "Complementary Exposure" pattern: the targets of the drugs both hit the HCoV-host subnetwork, but target separate neighborhoods in the human interactome network. In summary, this study offers powerful network-based methodologies for rapid identification of candidate repurposable drugs and potential drug combinations targeting 2019-nCoV/SARS-CoV-2.

Keywords: Bioinformatics; Comparative genomics; Proteomic analysis.

Conflict of interest statement

Conflict of interestThe authors declare that they have no conflict of interest.

© The Author(s) 2020.

Figures

Fig. 1. Overall workflow of this study.

Our network-based methodology combines a systems pharmacology-based network medicine platform that quantifies the interplay between the virus–host interactome and drug targets in the human PPI network. a Human coronavirus (HCoV)-associated host proteins were collected from literatures and pooled to generate a pan-HCoV protein subnetwork. b Network proximity between drug targets and HCoV-associated proteins was calculated to screen for candidate repurposable drugs for HCoVs under the human protein interactome model. c, d Gene set enrichment analysis was utilized to validate the network-based prediction. e Top candidates were further prioritized for drug combinations using network-based method captured by the “Complementary Exposure” pattern: the targets of the drugs both hit the HCoV–host subnetwork, but target separate neighborhoods in the human interactome network. f Overall hypothesis of the network-based methodology: (i) the proteins that functionally associate with HCoVs are localized in the corresponding subnetwork within the comprehensive human interactome network; and (ii) proteins that serve as drug targets for a specific disease may also be suitable drug targets for potential antiviral infection owing to common protein–protein interactions elucidated by the human interactome.

Fig. 2. Phylogenetic analysis of coronaviruses.

a Phylogenetic tree of coronavirus (CoV). Phylogenetic algorithm analyzed evolutionary conservation among whole genomes of 15 coronaviruses. Red color highlights the recent emergent coronavirus, 2019-nCoV/SARS-CoV-2. Numbers on the branches indicate bootstrap support values. The scale shows the evolutionary distance computed using the p-distance method. b Schematic plot for HCoV genomes. The genus and host information of viruses was labeled on the left by different colors. Empty dark gray boxes represent accessory open reading frames (ORFs). c–e The 3D structures of SARS-CoV nsp12 (PDB ID: 6NUR) (c), spike (PDB ID: 6ACK) (d), and nucleocapsid (PDB ID: 2CJR) (e) shown were based on homology modeling. Genome information and phylogenetic analysis results are provided in Supplementary Tables S1 and S2.

Fig. 3. Drug-target network analysis of the HCoV–host interactome.

a A subnetwork highlighting the HCoV–host interactome. Nodes represent three types of HCoV-associated host proteins: targetgable (proteins can be targeted by approved drugs or drugs under clinical trials), non-targetable (proteins do not have any known ligands), neighbors (protein–protein interaction partners). Edge colors indicate five types of experimental evidence of the protein–protein interactions (see Materials and methods). 3D three-dimensional structure. b, c KEGG human pathway (b) and gene ontology enrichment analyses (c) for the HCoV-associated proteins.

Fig. 4. A discovered drug-HCoV network.

a A subnetwork highlighting network-predicted drug-HCoV associations connecting 135 drugs and HCoVs. From the 2938 drugs evaluated, 135 ones achieved significant proximities between drug targets and the HCoV-associated proteins in the human interactome network. Drugs are colored by their first-level of the Anatomical Therapeutic Chemical (ATC) classification system code. b A heatmap highlighting network proximity values for SARS-CoV, MERS-CoV, IBV, and MHV, respectively. Color key denotes network proximity (Z-score) between drug targets and the HCoV-associated proteins in the human interactome network. P value was computed by permutation test.

Fig. 5. A discovered drug-protein-HCoV network for 16 candidate repurposable drugs.

a Network-predicted evidence and gene set enrichment analysis (GSEA) scores for 16 potential repurposable drugs for HCoVs. The overall connectivity of the top drug candidates to the HCoV-associated proteins was examined. Most of these drugs indirectly target HCoV-associated proteins via the human protein–protein interaction networks. All the drug–target-HCoV-associated protein connections were examined, and those proteins with at least five connections are shown. The box heights for the proteins indicate the number of connections. GSEA scores for eight drugs were not available (NA) due to the lack of transcriptome profiles for the drugs. b–e Inferred mechanism-of-action networks for four selected drugs: b toremifene (first-generation nonsteroidal-selective estrogen receptor modulator), c irbesartan (an angiotensin receptor blocker), d mercaptopurine (an antimetabolite antineoplastic agent with immunosuppressant properties), and e melatonin (a biogenic amine for treating circadian rhythm sleep disorders).

Fig. 6. Network-based rational design of drug combinations for 2019-nCoV/SARS-CoV-2.

a The possible exposure mode of the HCoV-associated protein module to the pairwise drug combinations. An effective drug combination will be captured by the “Complementary Exposure” pattern: the targets of the drugs both hit the HCoV–host subnetwork, but target separate neighborhoods in the human interactome network. ZCA and ZCB denote the network proximity (Z-score) between targets (Drugs A and B) and a specific HCoV. SAB denotes separation score (see Materials and methods) of targets between Drug A and Drug B. b–d Inferred mechanism-of-action networks for three selected pairwise drug combinations: b sirolimus (a potent immunosuppressant with both antifungal and antineoplastic properties) plus dactinomycin (an RNA synthesis inhibitor for treatment of various tumors), c toremifene (first-generation nonsteroidal-selective estrogen receptor modulator) plus emodin (an experimental drug for the treatment of polycystic kidney), and d melatonin (a biogenic amine for treating circadian rhythm sleep disorders) plus mercaptopurine (an antimetabolite antineoplastic agent with immunosuppressant properties).