Lessons from the COVID-19 pandemic for advancing computational drug repurposing strategies

These authors jointly supervised this work: Jan Baumbach, Josch Konstantin Pauling.

LipiTUM, Chair of Experimental Bioinformatics, TUM School of Life Sciences, Technical University of Munich, Munich, Germany

Abstract

Responding  quickly to unknown pathogens is crucial to stop uncontrolled spread of  diseases that lead to epidemics, such as the novel coronavirus, and to  keep protective measures at a level that causes as little social and  economic harm as possible. This can be achieved through computational  approaches that significantly speed up drug discovery. A powerful  approach is to restrict the search to existing drugs through drug  repurposing, which can vastly accelerate the usually long approval  process. In this Review, we examine a representative set of currently  used computational approaches to identify repurposable drugs for  COVID-19, as well as their underlying data resources. Furthermore, we  compare drug candidates predicted by computational methods to drugs  being assessed by clinical trials. Finally, we discuss lessons learned  from the reviewed research efforts, including how to successfully  connect computational approaches with experimental studies, and propose a unified drug repurposing strategy for better preparedness in the case  of future outbreaks.

Main

The novel SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2)  pathogen has infected around 60 million people and caused more than a  million deaths worldwide (https://covid19.who.int/; as of November 2020). As a result, there is a need to find treatments  that can be applied immediately to reduce mortality or morbidity.

Repurposing existing drugs is a rapid and effective way to provide such treatments  by identifying new uses for drugs that have well-established  pharmacological and safety profiles1. Many drugs used to treat different diseases have already been successfully repurposed and approved for new indications2. While repurposing can be conducted at any point in drug development,  its greatest potential can be applied to drugs that are already approved3. In the case of the COVID-19 pandemic, it is a fast and cost-efficient approach to identify novel treatments4.

Recent studies have increasingly employed computational methods to  systematically predict new drug targets or drug repurposing candidates.  In contrast to experimental high-throughput screening, in silico  approaches are faster, lower-cost, and can serve as an initial filtering step for evaluating thousands of compounds. Thus, they are useful for  prioritizing drugs that warrant further evaluation and experimental  validation. This requires the application of suitable algorithmic  approaches to identify mechanisms relevant or specific to the disease4.

This Review discusses current in silico drug repurposing efforts for  COVID-19, followed by a discussion of the lessons learned from different perspectives (from data resources to the quality of predictions) and a  proposed unified strategy to improve the response in potential future  outbreaks. The covered studies employed standard drug repurposing  workflows and data-driven algorithms.

As new studies are published almost every day, it is not possible to provide a broad and  comprehensive overview of all repurposing studies. Hence, this Review  focuses on the computational methods for drug repurposing, their  application, availability and feasibility in a selection of studies  (peer-reviewed and preprint) that were selected to cover a wide variety  of different methods. It is worth noting that most of these studies are  not considered successful clinically. Nevertheless, it is important to  properly evaluate and improve the predictive power of in silico  approaches that are capable of utilizing information from existing drugs as well as host and virus biology, even with limited availability of  data on the novel emerging pathogen. This promotes a rapid and practical response to infection and therefore improves success in future  pandemics, particularly in tackling the rise in infection cases at the  early stages of the pandemic or ahead of vaccine development.

Data resources

Besides experimental datasets, the rapid availability of resources that  integrate different data types is crucial in a pandemic. Sharing data  accelerates research, as computational methods depend on high-quality  datasets, and experimental labs do not need to collect the information  on their own. The large number of resources used in COVID-19 drug  repurposing studies have shown that data can be quickly generated and  gathered through strong community efforts. This section presents a  selection of data resources used in the reviewed studies to describe the resource types that accelerated computational drug repurposing  approaches: most of them are general data resources that were already  established before the pandemic but that have been extended with  COVID-19 or SARS-CoV-2-specific data. The resources used in the reviewed studies are listed in Supplementary Table 1. A list of COVID-19 specific data resources that were not used in the  reviewed studies but may become relevant in the future is given in  Supplementary Table 2.

Molecular data resources

All molecular data used in the reviewed publications were extracted from  already established, general data resources that were quickly extended  with SARS-CoV-2-specific data. Resources such as GenBank5, the GISAID initiative6, or UniProt7 provide genomic/proteomic sequence information about hosts and  SARS-CoV-2. Structural resources collecting information about proteins,  such as the Protein Data Bank (PDB)8, were extended by various SARS-CoV-2-specific proteins. Finally,  transcriptome resources that collect gene expression data were used in  several COVID-19 drug repurposing approaches. For instance, the  Genotype-Tissue Expression (GTEx)9 program offers insights into tissue-specific gene expression.  Expression in lung tissues is of high interest in COVID-19 drug  repurposing research and was often integrated in computational models or studies. Other resources, such as the LINCS L1000 database10, profile gene expression changes under certain drug treatment conditions and were used to identify drugs with reverse expression profiles to the samples infected with SARS-CoV-2.

Network and interaction resources

Protein–protein interaction (PPI) networks enable visualization and analyses of the  interactions between either host or virus proteins and other host  proteins. Furthermore, PPI networks allow for particular adaptation and  search strategies (for example, edge filtering) and can be connected to  drug resources. Gordon et al.11 identified 332 high-confidence virus–host interactions between  SARS-CoV-2 and human proteins. It was the only newly created and  exclusively SARS-CoV-2-related resource used in the reviewed  publications of this work. VirHostNet12,13, a virus–host PPI resource that already existed before the 2019/2020  SARS outbreak, was expanded with 167 new SARS-CoV-2 interactions. In  contrast to virus–host PPIs, host PPIs are not virus specific. All  resources that were used in the reviewed studies were already available  before the pandemic but have since been widely used in COVID-19 drug  repurposing approaches14,15. Besides molecular networks, knowledge graphs, such as the Global Network of Biomedical Relationships (GNBR)16, have demonstrated their utility for drug repurposing. These networks  comprise various types of biological relationships assembled from  literature and were integrated into COVID-19 drug repurposing approaches17.

Drug and trial resources

Drug databases that already existed before the pandemic and that are  continuously extended with newly developed drugs were used to connect  the results of different approaches to potential drugs. A widely used  drug database is DrugBank18, with more than 13,000 drug entries of approved and in-trial drugs, including drug targets. On the other hand, ChEMBL19 and ZINC1520 contain millions of compounds that exhibit drug-like properties.

Drug repurposing approaches also benefited from trial databases as they can  be used to validate whether the predicted drugs are already in trial or  have not yet been evaluated. Examples of such resources are the EU  Clinical Trials Register (https://www.clinicaltrialsregister.eu/) and ClinicalTrials.gov (https://clinicaltrials.gov/). The latter contains more than 350,000 research studies from 219 countries.

Drug repurposing studies

Various clinical, experimental and computational drug repurposing  efforts have been rapidly mobilized prioritizing compounds to identify  promising drug candidates for the SARS-CoV-2 pandemic. In this section,  we examine a selection of studies representing the different  computational approaches to identify potential new targets and  repurposable drugs for COVID-19.

Virus-targeting approaches

Virus-targeting approaches mostly rely on structure-based drug screening methods, which take the three-dimensional structures of target proteins to predict  affinities or interaction energies of known chemical compounds to the  proteins (Fig. 1). These methods were mainly used to identify candidate drugs that target  viral proteins, so we refer to them as virus-targeting approaches,  although they can also be applied to host proteins. Two main  methodological workflows were applied, namely, structure-based21 and deep-learning (DL)-based drug screening. Here, we describe these methods and compare 23 COVID-19 drug repurposing studies22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44.

The input data consist of protein structure information (experimental or predicted) and chemical structure of drugs from public databases. Two  analysis workflows can be applied: standard analysis consisting of  docking followed by molecular dynamics (MD) simulations and DL-based  analysis. Finally, the output data of both approaches generally consist  of a ranking of drugs based on their (predicted) docking scores. The  drugs can be further evaluated by whether or not they are in clinical  trials.

Structure-based drug screening

The first step for structure-based screening is the selection of the drug  library and the target protein. For COVID-19, the intuitive candidate  for targeting virus proteins were antivirals. Thus, many studies limited their search to these. The number of screened drugs ranged from 3  (ref. 37) to 123 antiviral drugs33. Broader studies, such as that by Chen et al.26, combined compounds from the KEGG (Kyoto Encyclopedia of Genes and Genomes) and DrugBank databases to screen 7,173 drugs.

The other crucial step is the selection of the target protein and its  corresponding three-dimensional structure (experimental or predicted).  Wu et al.40 performed screening on 19 encoded proteins of the virus. By comparison, most other studies focused on the 3CLpro, envelope (E), spike, RNA  polymerase and methyltransferase proteins.

Virtual screening of the drug libraries utilized established software, such as Autodock45 and Glide46. Candidate drugs were selected using respective scoring methods, followed by validations with molecular dynamics simulations30,37.

Most drugs were predicted for 3CLpro (Supplementary Table 3), which was also the focus of most studies (17 studies), followed by RdRp and PLPro. For 3CLpro, the predictions ranged from 2 (ref. 29) to 27 (ref. 40) drugs per study. The 5 most frequently predicted drugs were ritonavir  (8 studies), lopinavir (6 studies), nelfinavir, remdesivir and  saquinavir (5 studies each). However, 99 of the candidate drugs were  only predicted in 1 study, showing a high variability in the resulting  candidate sets. Interestingly, the studies that screened full databases  also predicted antiviral drugs as top scorers (Supplementary Table 4). Of the 23 studies, 10 have not yet been peer-reviewed, which we discuss in the section on ‘A unified drug repurposing strategy’.

DL-based repurposing strategies

DL models can predict binding affinities or docking scores and have shown  advantages over conventional docking protocols. While standard docking  protocols are limited to millions47, DL approaches can analyze billions of chemical compounds. This allows  them to be applied to whole databases, which increases the diversity of  the tested compounds and the likelihood of finding unconventional  compounds47. Furthermore, they are capable of processing more (physico-)chemical features48 and can find features related to a non-favorable docking47. However, most of these methods require datasets for training, which  often come from real docking simulations; thus, the performance of many  DL-based approaches still rely on the accuracy of the docking software  used for training.

Ton et al.42 developed DeepDocking47, which utilizes quantitative structure–activity relationship models  trained to predict docking scores of compounds targeting the SARS-CoV-2  3CLpro protein. It requires fewer docking pipelines, since it performs  docking only on subsets of compounds and can produce a reduced list of  compounds, which is also enriched in potential top hits.

Nguyen et al.49 developed the method MathDL, which utilizes low-dimensional  mathematical representations of the drug–target protein complex  structures, which are then fed to DL algorithms to predict binding  energies of drug–protein complexes. For SARS-CoV-2, the authors used  experimental binding affinity data from SARS-CoV ligand–3CLpro complexes from PDBbind and SARS-CoV protease inhibitors as training data to  predict binding energies on DrugBank compounds for SARS-CoV-2 3CLpro  (ref. 50) and does not depend on docking software.

Beck et al.44 developed a DL-based drug-target interaction prediction model, named  Molecule Transformer-Drug Target Interaction. It utilizes simplified  molecular-input line-entry system (SMILES)51 representations for drugs and protein sequences as input for training  and predicts affinities. For SARS-CoV-2, the model was trained on  commercially available antiviral drugs and viral target proteins.  Antiviral drugs already used against SARS-CoV-2 were found among the  candidate drugs identified.

Host-targeting approaches

Host-targeting approaches involve identifying potential drugs that interfere with host mechanisms that contribute to viral pathogenesis, which also makes them less prone to drug resistance52,53. In addition, SARS-CoV-2 infections can trigger a hyper-reactive immune  response characterized by the excessive release of pro-inflammatory  cytokines and chemokines54. Thus, drugs that modulate the host immune response can benefit  critically ill patients with COVID-19 by targeting specific dysregulated pathways54,55,56.

Signature-based approaches

Signature-based approaches primarily utilize transcriptome datasets from samples  infected with SARS-CoV-2 or closely related human coronaviruses to  identify candidate drugs through connectivity mapping (Fig. 2), a well-established approach that relies on finding drug-induced  expression signatures exhibiting reverse profiles to a disease signature57,58. Several studies adopted this as a primary method for identifying new therapeutics for COVID-19. Loganathan et al.59 performed differential expression analysis of virus-infected cells and  extracted consistently dysregulated genes in infected conditions. They  were used to query the Connectivity Map database58 for drug perturbation profiles exhibiting anti-correlated expression  signatures. A modified approach was implemented by Jia et al.60, wherein expression data from infected and healthy individuals were used as input to a pathway-guided drug repurposing framework. They  identified disease co-expression clusters and performed enrichment  analyses prior to reverse signature matching60.

Signature-based methods involve finding drug-induced expression profiles that exhibit reverse patterns to the coronavirus disease signature.  Network-based approaches typically assemble heterogeneous networks from  diverse data types, including gene–disease associations or drug–target  associations. Algorithms such as network proximity, random  walk/diffusion-based methods, or network enrichment are then employed.  Some studies combined them with machine-learning-based methods,  particularly autoencoders and graph convolutional networks. The outputs  can be ranked lists of host targets or drug candidates.

Network-based approaches

The general network-based approach applied in drug repurposing studies on  COVID-19 integrates multiple data sources, including virus–host  interactions, PPIs, co-expression networks, functional associations or  drug–target interactions (Fig. 2). Network-based algorithms or topology measures are applied to the  assembled networks to identify relevant host protein targets or regions  of the host interactome that can be targeted.

Multiple studies implement random-walk-based algorithms as the primary method to identify new putative drug targets. Law et al.61 implemented several algorithms on a virus–host interactome to identify  additional SARS-CoV-2 interactors. The coronavirus spike protein  primarily has been established to mediate viral entry into host cells62. Similarly, but focusing on a specific context, Messina et al.63 explored the pathogenic mechanisms triggered by the spike protein using data from three closely related coronaviruses. They implemented a  random walk algorithm on assembled molecular networks using the spike  protein as seed to identify relevant targets for COVID-1963. In addition, CoVex64 implemented TrustRank65, a variant of the PageRank66 algorithm, to propagate scores from user-defined seeds to the other host proteins and rank host drug targets.

Network proximity relies on the principle that a drug can be effective if it  targets proteins within the neighborhood of disease-associated proteins  in the interactome67. Zhou et al.68 utilized this concept to compute the network proximity measure between  drug targets and coronavirus-associated proteins in the human  interactome. They also used the ‘complementary exposure’ pattern, which  is based on the shortest distance between targets of two drugs predicted by network proximity, to identify potential drug combinations to treat  COVID-19 patients68.

Several studies combined multiple network-based strategies to predict drug candidates. Gysi et al.69 characterized and extracted a COVID-19 disease module using  experimentally determined SARS-CoV-2 interactors. They performed  network-based analyses accounting for tissue specificity and potential  disease comorbidities. They employed a multi-modal approach to the  virus–host interactome integrating network proximity, diffusion state  distance and graph convolutional networks (GCNs) to identify drugs that  can perturb the activity of host proteins associated with the COVID-19  disease module. The final drug list was obtained by rank aggregation  from the different pipelines69.

CoVex64 is a web platform for exploring SARS-CoV and SARS-CoV-2 virus–host–drug interactomes64. Users can predict drug targets and drug candidates using several graph  analysis methods that allow custom seed proteins as input. For instance, KeyPathwayMiner70 is a network enrichment tool that identifies condition-specific  subnetworks by extracting a maximally connected subnetwork from the host interactome starting from the seeds. CoVex also implements a weighted  multi-Steiner tree method that aggregates several non-unique  approximations of Steiner trees, which are subnetworks of minimum cost  connecting the set of seeds, into a single subnetwork.

Other studies additionally utilize machine learning to predict drug candidates against SARS-CoV-2. Belyaeva et al.71 implemented a hybrid approach between signature matching and  network-based methods. Using autoencoders, they learned feature  embeddings for drugs using drug-induced expression profiles to identify  drugs exhibiting reverse profiles to the SARS-CoV-2 infection signature. Steiner tree and causal network discovery algorithms were then used to  extract the mechanisms mediated by both SARS-CoV-2 and aging71. Ge et al.72 constructed a virus-related knowledge graph and employed a GCN  algorithm. The list of drug candidates was further filtered for existing evidence of antiviral activities through text mining72. Similarly, Zeng et al.17 assembled a large-scale knowledge graph derived from PubMed articles. A GCN model was then applied to learn low-dimensional embeddings of the  nodes and edges17.

Lessons learned

In the following, we examine the quality and potential of the reviewed  data resources and computational methods in order to improve the  response in future pandemics.

Data resources

The availability of molecular datasets is a precondition to develop drug  repurposing methods quickly. Besides that, network-based resources were a large driver in drug repurposing. However, a large portion of the  publications are based on only a few primary resources, which always  induces the risk of bias or measurement errors. In addition, the only  type of molecular interaction network used was PPI. Still, high  confidence PPIs are needed since, for instance, none of the approaches  included structure data. In the future, other network types, such as  gene regulatory networks, should be considered. Other data resources,  such as off-label data for drugs, should also be integrated in drug  repurposing studies.

Finally, existing drug and trial resources  were widely used for developing the drug repurposing pipelines. However, we observed no standardization in trial resources, making it hard to  analyze trials for certain drugs due to different names, different  spellings, or typing errors. Standardization is usually implemented for  drug resources (for example, DrugBank), but some drugs undergoing trials could not be found in the databases. Keeping the resources up to date  and interconnected should be a focus and will enhance accessibility.

Computational predictions

Assessing the quality of predictions is challenging, since many studies are not  peer-reviewed, do not perform experimental evaluation, or rely on  clinical trial databases. We examined the quality of predictions by  determining the overlap between the final candidate drug lists from the  individual studies and the drugs undergoing clinical trials from  ClinicalTrials.gov (https://clinicaltrials.gov/) and Biorender (https://biorender.com/covid-vaccine-tracker) databases. In addition, we provide supplementary in vitro screening data, such as IC50 values for viral targets and inhibition indices from cell culture studies for SARS-CoV-2 (Supplementary Data 1). Our effort to compile these data shows that a substantial number of predictions have not been experimentally tested.

Evaluating virus-targeting approaches

We identified 53 drugs predicted with docking simulations that are undergoing current trials (Supplementary Table 5). Wu et al.40 identified most of the drugs (36 drugs); however, these drugs were  predicted for multiple viral proteins (for example, chlorhexidine for 11 and methotrexate for 6 different viral proteins). This indicates that  their approach did not yield specific and feasible candidates. After  excluding this study, the remaining drugs were only predicted for one  specific protein each, except for chloroquine (3CLpro and PLpro) and  remdesivir (3CLpro and RdRp). The top five drugs in clinical trials,  which were predicted by docking simulations using the 3CLpro main  protease, were predicted by 14.3% (darunavir), 19.0% (remdesivir), and  23.8% (lopinavir, nelfinavir, ritonavir) of the total number of included docking studies (Supplementary Table 6), showing that for each drug, the majority of studies were not able to  predict them. Similar drugs were identified by the DL approach of Beck  et al.44, who identified ritonavir, lopinavir and remdesivir, which are being  tested in multiple clinical trials. However, these antiviral drugs have  not yet shown well-defined results in patients. For ritonavir/lopinavir, only four trials are completed73,74,75,76 and preliminary results suggest no difference in the outcome after treatment77,78,79. Further investigation is required80. For remdesivir, some trials have been completed and the preliminary results in patients81,82,83 and human cell lines84 showed that it could be effective in treating SARS-CoV-2 infection.

Antiviral drugs are always the top hits among a large selection of drugs from  databases, indicating high accuracy of the methods. These drugs are good candidates for experimental screening or clinical trials, independently of how reliable the computational predictions are. More interesting  candidates are the additional drugs identified by these approaches;  however, little experimental validation is available for these drugs and the majority of them do not enter clinical trials. A similar situation  is observed in the emerging field of DL approaches, where most studies  focused on demonstrating the accuracy of their predictions and  developing benchmarking datasets85,86. DL and docking simulation-based approaches are promising tools to  identify repurposable drugs given their capacity to deliver results in a short time. While a standard workflow is already established for  docking simulations, DL-based approaches might robustly deliver testable candidate drugs. However, docking studies in particular were rarely  peer reviewed, found very different candidate sets and partially used  different scores for evaluation and ranking. This makes it necessary to  validate these results by systematic comparisons of experiments.

Evaluating host-targeting approaches

Host-targeting approaches typically involve integration and analysis of multiple omics types and employ data-driven network-based methods; thus, a major  limitation is the lack of gold-standard datasets and the scarcity of  data from the MERS-CoV (Middle East respiratory syndrome coronavirus)  and SARS-CoV outbreaks. Prior to the availability of sufficient  SARS-CoV-2-specific data, earlier studies utilized preliminary data or  augmented the analyses using data from closely related viruses. While  the quality of the predictions is highly data-dependent, continued  generation of SARS-CoV-2-specific omics data and pending results on  clinical studies are expected to improve the predictions. Clinical  expert knowledge remains crucial for filtering the drug predictions  based on criteria such as toxicity and pharmacological properties.  However, the efficacy of these candidate drugs in trial remains to be  established and firm conclusions cannot be made because of the limited  data availability.

The degree of overlap with drugs in clinical trials was generally low (Supplementary Tables 7 and 8), but more than half of the drugs (26 out of 41) predicted by an ensemble method primarily based on knowledge graphs17 are also undergoing clinical trials. While it should be noted that the  drugs registered for clinical trials were also used as their validation  set at the time of writing, more of their predicted drugs were  registered for clinical trials later on. We also noted several drugs  that were predicted by both signature-based and network-based approaches and thus warranted further examination (Supplementary Table 9). Ribavirin was predicted by four out of six studies17,60,69,71, thereby providing a mechanistic basis for its predicted efficacy.  Methotrexate, which is indicated for rheumatoid arthritis, was also  predicted by three studies17,68,69.

It is worth noting that several predicted compounds are currently used to  treat critically ill COVID-19 patients. An example is dexamethasone  (predicted by one signature-based60 and two network-based studies17,69), which was supported by the RECOVERY trial87. Hydrocortisone (predicted by three studies17,68,69) has also demonstrated efficacy for critically ill patients88. Dexamethasone and hydrocortisone are corticosteroids that act by  modulating an overactive immune response, which is typically observed in severely ill COVID-19 patients.

Notably, drugs reaching advanced  phases in clinical trials were not selected based on in silico  predictions, but were repurposed based on clinical experience with the  previous SARS or MERS outbreaks89 and selected based on known effects in alleviating disease symptoms.  Furthermore, the predictions were not followed-up by experimental  validation in the majority of the studies reviewed. This translational  gap between computational efforts for drug repurposing and clinical  application is a major and widely recognized bottleneck in drug  repurposing and medicine in general. Results from systematic validation  efforts will also be important for identifying the algorithms and  datasets that are specifically suitable for drug repurposing in the  COVID-19 context. Given the urgency of identifying effective therapies  in a pandemic, close collaboration between clinicians, experimental  biologists and computational biologists is expected to address this gap.

A unified drug repurposing strategy

Although overlaps between computationally predicted drug repurposing and  clinical trials exist, there are no indications that clinical trials  were conducted based on computational predictions, despite their  promising potential. For future pandemics, computational tools should be able to deliver promising sets of candidates, which could then be  validated in trials or screenings. Therefore, a unified strategy is  necessary. In the following, we identify important issues and discuss  potential solutions to make computational drug repurposing more  effective.

Availability of standardized data

Newly developed methods often rely on the same data types (Fig. 3a). The fast generation of different kinds of data in future disease  outbreaks is a key initial step. Notable examples are the interaction  data from Gordon et al.11 and the publication of the 3CLpro90 structure, which were both used by many subsequent studies. However,  experimental replication of datasets obtained from different  laboratories and the integration of different data types are crucial to  increase robustness and require improvement.

Fig. 3: Proposed elements of a unified drug repurposing strategy.

a, Availability of standardized data. b, Accessible workflows for computational predictions. c, Combination of predictions from different methods. d, Feedback from clinical experts of drug candidate sets and screening parameters. e, Validation of predicted drugs with different approaches.