Source-linked AI summary
Repurpose Open Data to Discover Therapeutics for COVID-19 using Deep Learning
Xiangxiang Zeng, Xiang Song, Tengfei Ma, Xiaoqin Pan, Yadi Zhou, Yuan Hou, Zheng Zhang, George Karypis, Feixiong Cheng
TL;DR
COVID-19 lacked proven effective medications, motivating drug repurposing approaches. The paper develops CoV-KGE, an integrated knowledge-graph deep-learning methodology, and identifies 41 repurposable drug candidates validated with infected-cell omics and ongoing clinical-trial data. The authors present the method as a prioritization tool for further investigation, not as a recommendation of specific drugs.
Problem
COVID-19 lacked proven effective medications, while clinical translation and preclinical animal studies were constrained.
Method
CoV-KGE integrates knowledge-graph deep learning with large-scale biological relationships to identify and validate repurposable COVID-19 drug candidates.
Results
The study identified 41 repurposable drug candidates using transcriptomic, proteomic, and ongoing clinical-trial data.
Takeaways & Limitations
CoV-KGE can prioritize existing drugs for further investigation and potential COVID-19 treatment development.
Takeaways & Limitations
Chloroquine and hydroxychloroquine had adverse effects including QT prolongation, and their safety and efficacy required testing in large-scale randomized controlled trials.
Abstract
from arXiv · showhide
There have been more than 850,000 confirmed cases and over 48,000 deaths from the human coronavirus disease 2019 (COVID-19) pandemic, caused by novel severe acute respiratory syndrome coronavirus (SARS-CoV-2), in the United States alone. However, there are currently no proven effective medications against COVID-19. Drug repurposing offers a promising way for the development of prevention and treatment strategies for COVID-19. This study reports an integrative, network-based deep learning methodology to identify repurposable drugs for COVID-19 (termed CoV-KGE). Specifically, we built a comprehensive knowledge graph that includes 15 million edges across 39 types of relationships connecting drugs, diseases, genes, pathways, and expressions, from a large scientific corpus of 24 million PubMed publications. Using Amazon AWS computing resources, we identified 41 repurposable drugs (including indomethacin, toremifene and niclosamide) whose therapeutic association with COVID-19 were validated by transcriptomic and proteomic data in SARS-CoV-2 infected human cells and data from ongoing clinical trials. While this study, by no means recommends specific drugs, it demonstrates a powerful deep learning methodology to prioritize existing drugs for further investigation, which holds the potential of accelerating therapeutic development for COVID-19.
Introduction
The study introduces CoV-KGE, an integrated knowledge-graph deep-learning methodology for rapidly identifying repurposable COVID-19 drugs. It uses large-scale publication-derived relationships and systematic molecular and clinical validation to prioritize candidates for further testing.
- Method: CoV-KGE uses a knowledge graph derived from 24 million PubMed publications and DrugBank to represent drug, disease, gene, and related biological relationships.The graph contains 15 million edges across 39 relationship types.
- Implications: The approach is intended to prioritize existing drugs for further COVID-19 testing rather than recommend specific treatments.The paper proposes large-scale randomized controlled trials as a next testing step.
- Method: The method learns low-dimensional representations of entities and relationships with the RotatE knowledge-graph embedding model.Relationships are modeled as rotations from source entities to target entities in vector space.
- Results: Using ongoing COVID-19 clinical-trial data for validation, CoV-KGE achieved AUROC = 0.85 for identifying repurposable drugs.The model selected the top 100 relevant drugs before downstream validation.
- Results: The study obtained 41 high-confidence repurposable drugs by combining predicted scores, clinical evidence, and transcriptomic and proteomic validation.Eleven identified drugs were or had been under clinical trials for COVID-19.
Figure Legends
CoV-KGE is a network-based deep-learning methodology that uses a large biomedical knowledge graph to prioritize COVID-19 drug candidates and evaluates them against molecular and clinical-trial evidence.
- Workflow: CoV-KGE uses RotatE in DGL-KE to prioritize high-confidence COVID-19 drug candidates from a knowledge graph.The graph contains relationships connecting drugs, diseases, genes, pathways, and expressions.
- Validation: Candidate drugs were prioritized using Amazon supercomputing resources and subsequently evaluated with transcriptomic, proteomic, and ongoing clinical-trial data.The validation data included gene-expression datasets from infected human cells, proteomic data from SARS-CoV-2-infected human cells, and COVID-19 trial data.
- Drug landscape: The learned two-dimensional drug representations group 14 drug types by first-level ATC codes, with semantically similar drugs mapped nearby.The representations were generated using t-SNE.
- Proposed mechanisms: A proposed treatment landscape combines antiviral candidates such as toremifene or niclosamide with anti-inflammatory agents such as melatonin or indomethacin.The passage presents this combination as a promising approach that requires further testing in randomized controlled trials.
Figure 3
Figure 3 summarizes selected CoV-KGE drug candidates and reports molecular activity measurements for toremifene, indomethacin, and niclosamide against coronaviruses.
- Antiviral candidates: Toremifene is presented as a selective estrogen receptor modulator with measured activity against Ebola virus, MERS-CoV, SARS-CoV-1, and SARS-CoV-2.Its reported values include IC50 = ~1 μM for Ebola, EC50 = 12.9 μM for MERS-CoV, EC50 = 11.97 μM for SARS-CoV-1, and IC50 = 3.58 µM for SARS-CoV-2.
- Anti-inflammatory candidates: Indomethacin is described as a COX1/2 inhibitor with IC50/EC50: 1 μM against SARS-CoV-2 and in vivo anti-SARS-CoV-2 activity in dogs.It is classified as a nonsteroidal anti-inflammatory drug.