Exploring biomedical literature and clinical trial records to identify new drug combinations for COVID-19 treatment

In a study published in the latest issue of the journal Pharmaceutics, researchers extensively searched publicly available severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) biomedical literature to identify drug combinations that could effectively treat coronavirus disease 2019 (COVID-19).

Drug repurposing accelerates the development of existing drug(s) as a treatment for emerging new diseases, such as COVID-19. To this end, abundant SAR-CoV-2 research publications are now available to help researchers identify COVID-19 repurposing drugs; and it is possible to validate these associations from evidence extracted from clinical trials designed to study these drugs.

Several past studies have used artificial intelligence, including machine learning, to accelerate COVID-19 drug repurposing. The researchers of the present study also identified drugs and protein targets from SARS-CoV-2 biomedical literature using a computational model. One of their key findings was that sometimes it might be crucial to use more than one drug for COVID-19 treatment.

About the study

In the present study, they tested a hypothesis that a combination of the Food and Drugs Administration (FDA)-approved drugs may be considered as potential COVID-19 drug candidate if it is present in SARS-CoV-2 biomedical literature and clinical trial-based evidence validates its combination with other drugs.

They developed a novel algorithm that found the intersection of the results from SARS-CoV-2 biomedical literature and clinical trials, constituting a candidate COVID-19 treatment, to form a map structure termed ‘clique’. In this way, they established that the drugs identified from both these sources were strongly associated within the source literature; a shred of consistent evidence for further testing.

They followed a four-step methodology: (1) drug name extraction using the Chemical Entities of Biological Interest (ChEBI) ontology, (2) network construction following association analysis, (3) clique detection, comparing and mapping similar cliques to find an intersection and (4) validation against clinical trials.

They sourced work experiments from two datasets extracted from ClinicalTrials.gov and PubMed, running the search query ‘COVID-19’. The cutoff date for the study was June 25, 2021, before which the search fetched a set of COVID-19-related clinical trial records and many more (110,000) SARS-CoV-2-related biomedical publications.

They analyzed the biomedical publications for drug names and their associations, the intervention and treatment sections of clinical trial records validated the drug associations identified in these publications. Finally, they presented final drug combinations to domain experts for interpretation.

Findings

The dataset of publications was huge and thus analyzed in smaller subsets, each comprising of 7,000 biomedical publications. In the network constructed from the medical publications, there were cliques of size two to six; contrastingly, there were cliques of size two to four in the network derived from the clinical trials.

When analyzed as one fold, 5,578 clinical trial records returned a total of 92 cliques, of which 78 were size two, 10 of size three, and 6 of size four which eliminated 3,551 cliques detected from the publications, and the comparisons identified drug matches of a maximum of three of the four components.

Among the two-drug combinations, seven drug pairs were possible to combine without any potential clinical issues, and included the following:

i) ruxolitinib (janus kinase inhibitor) and colchicine (anti-gout);

ii) hydroxychloroquine (antimalarial) and favipiravir (antiviral);

iii) azithromycin (macrolide antibiotic) and ivermectin (anthelmintic);

iv) hydroxychloroquine (antimalarial) and doxycycline (tetracycline antibiotic);

v) Daclatasvir (antihepaciviral) and sofosbuvir (nonstructural protein 5B (NS5B) nucleoside polymerase inhibitor).

Two drug combinations are already commercialized, including the lopinavir (protease inhibitor) and ritonavir (protease inhibitor) combination that exists as an FDA-approved medication under the brand name Kaletra; and another one is the nirmatrelvir/ritonavir combination, marketed as Paxlovid.

In addition, the work identified a few three-way drug combinations using the Search-n-Match algorithm. Of these, the drug combination of hydroxychloroquine, lopinavir, and favipiravir was worthwhile to investigate.

Conclusions

The study used a computational framework to validate the tested hypothesis and showed that it could be a promising COVID-19 drug repurposing prediction tool; additionally, after validating the drug combinations with evidence mined from clinical trial records, the authors further validated the identified clique patterns from literature-based networks by domain experts and categorized them based on combinability.

Overall, the study findings showed that FDA-approved drugs (in pairs or triples) are promising as COVID-19 treatments, such as the already commercialized nirmatrelvir/ritonavir marketed under the name Paxlovid.

The authors also cautioned that since they screened study-identified drug names and combinations using publicly available SARS-CoV-2 biomedical publications and ongoing COVID-19 clinical trials, these results must be validated further before usage in real-world settings except those commercialized, such as Kaletra.

In the future, studies should specifically identify more validation sources to improvise the understandings of the cliques that did not have corresponding coverage in the clinical trial records. These studies should also explore how such drug combinations may affect the treatment of COVID-19 patients with preexisting health conditions, such as asthma, depression, diabetes, and hypertension.

Written by

Neha Mathur

Neha is a digital marketing professional based in Gurugram, India. She has a Master’s degree from the University of Rajasthan with a specialization in Biotechnology in 2008. She has experience in pre-clinical research as part of her research project in The Department of Toxicology at the prestigious Central Drug Research Institute (CDRI), Lucknow, India. She also holds a certification in C++ programming.