About the COVID-19 Therapeutic Information Browser

Update (April 12, 2021)

A briefing presenting an analysis of published COVID-19 drug research and additional information about the NLP methods is available for download.

Download the Briefing

Introduction

Browser recommendation: Internet Explorer is not supported by this application. Please consider using an updated version of Chrome, Edge, Firefox, or Safari.

This site provides browsable therapeutic and vaccine-related information about SARS-CoV-2 (COVID-19) and other viruses. This information was extracted with natural language processing (NLP) from more than 100k PubMed, MedRxiv, and BioRxiv abstracts and clinicaltrials.gov summaries [1]. In addition, some of the information has been reviewed and curated by subject matter experts. Information on this site will be updated with results from new papers every few days.

The summary tables show the number of papers with information about a given virus and drug (Viral Therapeutics page), vaccine type (Vaccine Type page), or entry receptor (Entry Receptors page). For drugs and vaccines, the numbers of papers for selected research stages, e.g. clinical trials, are shown separately. Click on a number to see a listing of papers where this information was found; paper titles will be displayed below the summary table in a paper table. The columns of both tables are sortable by clicking on the column headings. The drug names and entry receptors at the top of the paper tables have hyperlinks to information in the DrugBank Open Data Vocabulary and UniProt respectively.

The paper tables contain information about the research stage of each paper, which comes from either NLP or expert curation as indicated. Some of the entries also include expert curated information about the efficacy of drugs as reported in the abstracts. This is based solely on the results and conclusions as stated in the abstracts; curators have not assessed the scientific merit of these claims.

For more information about curation's role and term definitions, please see the tabs under the “About” menu in the navigation bar to the left.

CAVEAT: Machines are now great at processing documents at-scale but currently are not nearly as accurate as human subject matter experts. Consequently, you will see mistakes in raw machine-extracted information posted here. Longer term, the curated results may be used to improve the accuracy and coverage of the NLP extractions.

This is a beta version prototype. Your feedback is welcome. Please send comments, suggestions, or bugs to covid-browser@groups.mitre.org.

[1]: The natural language processing (NLP) uses REACH reading software developed by the University of Arizona (M. Valenzuela-Escárcega et al, Large-scale automated machine reading discovers new cancer-driving mechanisms, Database, Volume 2018, 2018, bay098) that was adapted for virus-related information by MITRE. Data sources include: NCBI, clinicaltrials.gov, the CORD-19 Open Research Dataset, UniProt, and DrugBank.

Curation's Role

Our approach integrates natural language processing (NLP) with expert curation to improve the accuracy of the information presented. Our NLP pipeline generates results about viruses, drugs, vaccine types, entry receptors, and research stages. These results are sent to a MITRE-developed curation user interface, where subject matter expert curators can easily review the extracted information with text evidence. Curators determine whether a paper is about the virus and drug/vaccine type/entry receptor identified by NLP. If the curator determines that it is, the curator selects supporting sentence(s) of interest from the abstract to show in the Information Browser and the paper is marked as curated. When a curator determines that a paper is not about the extracted virus and drug/vaccine type/entry receptor, the paper is no longer shown for that result in the Information Browser.

Our NLP system identifies the research stages; human curators can also curate this information.Papers with research stage information from NLP only are marked “Machine-only” while papers with research stage information that has been reviewed or added by curators are marked “Curated”.

All the information about reported drug efficacy is added by curators. This is solely an assessment of what the abstract states or claims about the drug's efficacy in the study. Our curators do not assess the scientific merits of these results and conclusions.

Term Definitions

Research Stages

Only one research stage is listed for a virus and drug or vaccine per paper and the most advanced research stage reported in the abstract is preferentially selected. Research stages are listed below in ascending order from least to most advanced.

  • AI/in silico: Drug or vaccine has been designed, modeled, or predicted for activity against the virus using AI/in silico methods.
  • In vitro: Drug or vaccine has been studied/tested with purified proteins/enzymes/components and not in cells.
  • Cell line: Drug or vaccine has been screened/tested for activity against the virus in a cultured cell line(s).
  • Small animal: Drug or vaccine has been screened/tested for safety and prophylactic or therapeutic activity against the virus- in vivo in a small animal model(s) such as mice, ferrets, rabbit
  • Nonhuman primate: Drug or vaccine has been screened/tested for safety and prophylactic or therapeutic activity against the virus in vivo in a non-human primate (NHP) model(s).
  • Emergency use or case report: Reported use of a drug or vaccine for prophylactic or therapeutic treatment against the virus in vivo in human(s).
  • Clinical study: Drug or vaccine has been or is currently undergoing evaluation for safety and efficacy in humans in controlled, randomized Phase 1 – 4 clinical trial(s). Also includes observational studies, retrospective studies, and clinical meta-analyses where the drug plays a prominent role.
  • Review: Drug or vaccine research or testing progress is summarized in a review article, multiple research stages may be described (e.g., small animal, non-human primate, and clinical trials).

Drug Efficacy

  • Not effective: Reported testing results indicate that the drug did not display activity/efficacy against the virus at a given research stage(s).
  • Mixed data/inconclusive: Reported testing results indicate that the drug displayed inconsistent/mixed activity/efficacy against the virus at a given research stage(s).
  • Effective: Reported testing results indicate that the drug displayed activity/efficacy against the virus at a given research stage(s).
  • No data: Drug has been or is currently undergoing evaluation/testing for activity/efficacy against the virus, but no effectiveness data has been reported. This includes active clinical trials and authorizations/protocols for emergency use.
  • Not tested: Drug has not been or is not being evaluated for activity/efficacy against the virus. Drug activity/efficacy against the virus is unknown but the paper may present hypotheses about this without results.

Vaccine Types

  • Inactivated: Includes vaccines where the virus has been inactivated by heat, chemical or radiation methods.
  • Live attenuated: Includes vaccines where the targeted virus has been attenuated by passage, genetic modification, or natural reassortment with a non-pathogenic virus (in vivo or in vitro). This category does not include vaccines where select virus antigens are engineered for expression in a different virus backbone (see viral vector category below); these can be referred to as “live, attenuated chimeric” vaccines in literature but should be categorized as viral vector vaccine.
  • Protein subunit: Includes any vaccines that use select virus proteins as an antigen including purified proteins from a virus; recombinant proteins; recombinant epitopes; and Virus-like particles (VLP) or nanoparticles.
  • Virus-like particles (VLP)/nanoparticles: include 1) when discreet protein(s) self-assemble or are encouraged to assemble into particles; and 2) Virosomes. This technology involves purification or expression of discreet virus proteins that are reassembled with phospholipids to form a VLP that resembles a virus particle without genomic material. The protein subunit category excludes vaccine that used target virus proteins expressed in a viral vector.
  • Viral vector: Includes any vaccine engineered so that selected antigens from the target virus are expressed with the backbone of a different virus; both standard virus vectored vaccines as well as engineered chimeric viruses (can be referred to as “live attenuated chimeric”) are included in this category.
  • DNA vaccine: Includes DNA constructs engineered to express selected virus antigen(s).
  • RNA vaccine: Includes RNA constructs engineered to express a selected virus antigen(s).
  • Vaccine type not identified: Indicates a paper is about vaccine research but is either not about a particular vaccine type or NLP did not identify a particular vaccine type. This includes early vaccine-related research, such as research on epitopes.

This figure shows the numbers of papers found with information about pairs of drugs and viruses. Click on a number in the figure to see the papers below the figure. Click on the numbers between Previous and Next to see additional drugs and papers.

Loading...

This figure shows the numbers of papers found with information about pairs of vaccines and viruses. Click on a number in the figure to see the papers below the figure. Click on the numbers between Previous and Next to see additional vaccines and papers.

Loading...

This figure shows the numbers of papers found with information about pairs of entry receptors and viruses. Click on a number in the figure to see the papers below the figure. Click on the numbers between Previous and Next to see additional entry receptors and papers.

Loading...