Drug Repurposing in the AI Era: Strategies and Success Stories

What Is Drug Repurposing in the AI Era?

Drug repurposing (also called drug repositioning) is the identification of new therapeutic indications for existing drugs. In the AI era, this process has evolved from serendipitous discovery to systematic, data-driven prediction. AI methods—including network medicine, transcriptomic signature matching, and knowledge graph reasoning—can identify non-obvious connections between approved drugs and diseases.

The advantage is compelling: repurposed drugs already have established safety profiles, manufacturing processes, and pharmacokinetic data, potentially reducing development time from 10-15 years to 5-8 years and cost from $2.6 billion to $300 million. AI-driven repurposing was prominently validated during the COVID-19 pandemic, when baricitinib was identified by BenevolentAI's knowledge graph platform and subsequently received FDA emergency use authorization.

Data: Drug Repurposing Impact

How: AI-Driven Drug Repurposing Pipeline

Step 1: Data Integration

1. Build comprehensive knowledge graph:
   • Drug-target interactions (DrugBank, ChEMBL)
   • Disease-gene associations (DisGeNET, OMIM)
   • Protein-protein interactions (STRING, BioGRID)
   • Pathway data (KEGG, Reactome)
   • Clinical trial data (ClinicalTrials.gov)
   • Scientific literature (PubMed, NLP extraction)
2. Integrate multi-omics data:
   • Transcriptomic signatures (LINCS L1000, Connectivity Map)
   • Proteomics, metabolomics
   • Electronic health records

Step 2: AI-Powered Prediction

1. Network medicine: Compute network proximity between drug targets and disease genes
2. Transcriptomic matching: Compare drug-induced gene expression changes with disease signatures (L1000, CMap)
3. Knowledge graph reasoning: Graph neural networks predict drug-disease edges
4. Literature mining: NLP identifies drug-disease associations in text
5. Multi-modal fusion: Combine all approaches with ensemble learning

Step 3: Candidate Prioritization

1. Rank candidates by:
   • Predicted therapeutic effect
   • Safety profile (ADMET, contraindications)
   • Commercial viability (patent status, market size)
   • Mechanistic plausibility
2. Apply domain expertise to filter predictions
3. Select top candidates for experimental validation

Step 4: Experimental Validation

1. In vitro testing in disease-relevant cell models
2. In vivo proof-of-concept in animal models
3. Biomarker confirmation
4. Dose optimization for new indication

Step 5: Clinical Development

1. Repurposing trial design (often Phase II directly)
2. Regulatory pathway:
   • FDA: 505(b)(2) application (new indication for approved drug)
   • EMA: Type II variation or new marketing authorization
3. IP strategy (method-of-use patents, formulations)

Comparison: AI Repurposing Success Stories

Summary: Key Takeaways

1. Drug repurposing reduces development time (3-8 years vs. 10-15) and cost ($300M vs. $2.6B).
2. AI-driven repurposing uses knowledge graphs, transcriptomic matching, and network medicine.
3. Baricitinib for COVID-19 (BenevolentAI) is the landmark AI repurposing success.
4. The repurposing market is projected to double by 2030, driven by AI and rare disease focus.
5. Regulatory pathways (FDA 505(b)(2), EMA variations) are established for repurposed drugs.

References

1. Pushpakom, S. et al. "Drug repurposing: progress, challenges and recommendations." Nature Reviews Drug Discovery 18, 41-58 (2019).
2. Richardson, P. et al. "Baricitinib as potential treatment for 2019-nCoV acute respiratory disease." The Lancet 395, e30-e31 (2020).
3. Jarada, T.N. et al. "Drug repurposing through network-based approaches." Network Biology (2023).
4. FDA. "Orange Book: Approved Drug Products with Therapeutic Equivalence Evaluations." (2024).
5. Corsello, S.M. et al. "Discovering the anti-cancer potential of non-oncolytic drugs by systematic toxicity profiling." Nature Cancer 1, 235-248 (2020).