Data-driven strategies for drug repurposing: insights, recommendations, and case studies.
Level 5 - mechanism / opinion, no new human data
In silico bioinformatic database analysis and computational modeling (graded by design analogy, non-clinical)
PubMed 41283811 · doi:10.1093/bib/bbaf625
What was done
The authors compared drug-target interaction data across three curated databases (ChEMBL, BindingDB, and GtoPdb), evaluating coverage of compounds and targets, curation approaches, and release histories. They classified ChEMBL targets into 12 high-level biological families, mapped 817 clinically approved drug indications into 28 broad therapeutic groups, and profiled drug physicochemical properties across these categories. Additionally, they deployed a pathway-based computational pipeline to predict repositioning candidates among FDA-approved drugs across 10 cancer types.
What was found
The authors mapped 817 approved drug indications into 28 therapeutic groups and identified links between physicochemical properties and specific therapeutic categories. Cross-indication approvals were analyzed to highlight areas of high repurposing potential, and pathway-based candidates were predicted for 10 cancer types. Specific performance metrics, statistical values, and specific repurposing hit rates were not reported in the abstract.
Why it matters
The study provides a consolidated data-driven framework and comparative benchmark of major bioinformatic resources to guide compound prioritization and pathway-based computational drug repurposing.
Limits
The study relies entirely on in silico analysis of curated databases without reported in vitro, in vivo, or clinical validation of predicted repurposing candidates. The abstract does not report quantitative performance benchmarks, validation metrics, or error rates for the prediction pipeline.
Cited by
- context There are approximately 4,000 FDA-approved drugs approved for about 4,000 human diseases.