Owing to a lack of the complete drug-target information (such as the molecular “promiscuity” of drugs), the dose–response and dose–toxicity effects for both repurposable drugs and drug combinations cannot be identified in the current network models. For example, Mesalazine, an approved drug for inflammatory bowel disease, is a top network-predicted repurposable drug associated with HCoVs (Fig. 5a). Yet, several clinical studies showed the potential pulmonary toxicities (including pneumonia) associated with mesalazine usage81,82. Integration of lung-specific gene expression23 of 2019-nCoV/SARS-CoV-2 host proteins and physiologically based pharmacokinetic modeling83 may reduce side effects of repurposable drugs or drug combinations. Preclinical studies are warranted to evaluate in vivo efficiency and side effects before clinical trials. Furthermore, we only limited to predict pairwise drug combinations based on our previous network-based framework28. However, we expect that our methodology remain to be a useful network-based tool for prediction of combining multiple drugs toward exploring network relationships of multiple drugs’ targets with the HCoV–host subnetwork in the human interactome. Finally, we aimed to systematically identify repurposable drugs by specifically targeting nCoV host proteins only. Thus, our current network models cannot predict repurposable drugs from the existing anti-virus drugs that target virus proteins only. Thus, combination of the existing anti-virus drugs (such as remdesivir64) with the network-predicted repurposable drugs (Fig. 5) or drug combinations (Fig. 6) may improve coverage of current network-based methodologies by utilizing multi-layer network framework16.