# AI for Drug Repurposing — Build Mechanism + Clinical Evidence Chains > **Reviewed:** 2026-05-28 > **Canonical HTML:** https://bioskepsis.ai/use-cases/ai-for-drug-repurposing > **Publisher:** BioSkepsis (EFEVRE TECH LTD, Larnaca, Cyprus) ## Goal Build a complete mechanistic and clinical evidence chain supporting a new therapeutic indication for an existing drug by progressively drilling from broad overview to specific sub-questions. Each step refines the evidence — molecular rationale, clinical outcomes, mechanistic nuances, comparative effectiveness, and patient stratification. ## Built for translational researchers and pharma R&D - **Pharma R&D scientists:** map the literature foundation for life-cycle management, label-expansion, and Phase IIb/III repurposing decisions. - **Biotech research teams:** build mechanism + clinical evidence chains for licensing discussions, pipeline pivots, and new indication strategies. - **Academic translational researchers:** ground hypothesis-driven repurposing work in verified mechanistic literature with traceable citations. ## The 7-step workflow — drug-indication pair to patient stratification ### Step 1 — Enter a focused research question about a specific drug-indication pairing Frame the question around what published evidence supports a known drug or drug class being effective for a new therapeutic use. Specify the drug (or drug class) and the candidate indication clearly. **Example query** > "What is the published mechanistic and clinical evidence that aspirin has therapeutic potential in colorectal cancer prevention?" ### Step 2 — Review the initial AI synthesis BioSkepsis returns a structured answer covering the molecular rationale for the repurposed indication and the clinical evidence organised by sub-indication or disease stage. Pay attention to the **unverified citations panel** — this is where the system catches misquoted effect sizes or misattributed findings from the source literature. ### Step 3 — Drill into the core mechanism of action in the new indication Use a suggested follow-up or type your own question to explore how the drug's known pharmacology engages the pathways relevant to the new disease. Ask how the mechanism behaves under different biological states (e.g., healthy vs. diseased tissue, naive vs. chronic exposure). **Example query** > "What molecular mechanisms distinguish aspirin's effects on COX-2 inhibition in colorectal epithelium versus its systemic anti-platelet activity?" ### Step 4 — Link the mechanism to a measurable clinical or behavioural outcome Ask how the mechanistic findings from Step 3 correlate with observable endpoints. This tests whether the biology translates into something that a clinical trial could measure. **Example query** > "How do these mechanistic findings on prostaglandin signaling correlate with observed reductions in colorectal adenoma incidence in clinical trials?" ### Step 5 — Explore a nuanced distinction that affects therapeutic design Identify a conceptual split that matters for how the drug would be developed for the new indication (e.g., motivation vs. pleasure, acute vs. chronic pathology, upstream vs. downstream targets) and ask BioSkepsis to dissect which side the evidence supports. **Example query** > "Does aspirin's chemopreventive effect selectively target sporadic versus Lynch syndrome-associated colorectal cancer?" ### Step 6 — Benchmark against the current standard of care for the new indication Ask how the repurposed drug compares to existing approved treatments on specific metrics — effect sizes, hazard ratios, hospitalisation rates, or adverse event profiles. **Example query** > "How does aspirin's chemopreventive effect on colorectal cancer compare to established screening strategies in real-world cohorts?" ### Step 7 — Identify a patient stratification factor for the new indication Ask whether a specific patient characteristic (e.g., BMI, genotype, comorbidity, disease severity) predicts who responds to the repurposed therapy. This surfaces both positive predictors and any paradoxical or adverse findings in subgroups that would shape trial design. **Example query** > "What role does baseline body mass index or PIK3CA mutation status play in predicting aspirin's chemopreventive efficacy?" ## What you walk away with - **Mechanistic rationale chain** — from known pharmacology to disease-relevant pathways - **Clinical evidence organised by sub-indication** — population, dose, comparator, endpoints - **Comparative effectiveness benchmarks** — effect sizes vs. standard of care - **Patient stratification factors** — positive predictors and paradoxical subgroups ## Try the live workflow Pre-loaded BioSkepsis session demonstrating the full drug-repurposing workflow: **https://app.bioskepsis.ai/research/published-mechanistic-and-clinical-evidence-that-glp-1/71DcET-4_5DUp-5U8ajyAg** ## Honest limits - **AI cannot propose unpublished candidates.** BioSkepsis surfaces mechanistic links across the published literature; it will not invent novel drug-target pairings. - **AI does not replace clinical judgement.** Translation from preclinical mechanism to trial design requires regulatory, statistical, and ethics expertise. - **AI does not guarantee a clean evidence chain.** Where studies disagree, BioSkepsis surfaces the disagreement rather than smoothing it into false consensus. - **Final review remains human.** Every claim should be verified against the cited source before inclusion in a regulatory filing. ## FAQ **What kinds of drug repurposing questions does BioSkepsis support?** Any biomedical drug-indication pairing supported by published literature. Common workflows include CNS-active drugs repurposed for addiction or neurodegeneration, metabolic drugs repurposed for inflammation or cardiovascular disease, oncology drugs repurposed across cancer types, and antivirals repurposed for autoimmune disease. **Can BioSkepsis identify novel drug repurposing candidates I have not considered?** BioSkepsis surfaces mechanistically related drugs and pathways across its biomedical knowledge graph. It will not propose unpublished candidates but will reveal published mechanistic links you may not have known about. **How does BioSkepsis distinguish baseline predictors from post-treatment correlates?** The Step 5 nuance probe is built for exactly this. Ask whether a biomarker was measured pre-treatment as a predictor or post-treatment as a correlate, and BioSkepsis distinguishes the two and flags any study design that conflates them. **Can the output be used for an IND or grant submission?** BioSkepsis output is structured to map into preclinical rationale sections. Final IND or grant submission requires regulatory and PI review. **Does BioSkepsis cover pharmacology and clinical trial literature?** Yes. The corpus includes pharmacology, pharmacokinetics, clinical trial reports, real-world evidence, comparative effectiveness studies, and preclinical mechanistic literature. ## Related - [Use Case: AI for Grant Writing](/use-cases/ai-for-grant-writing) - [Use Case: AI for Precision Medicine](/use-cases/ai-for-precision-medicine) - [Blog: AI for Drug Repurposing](/blog/ai-for-drug-repurposing) --- BioSkepsis is a product of EFEVRE TECH LTD (Larnaca, Cyprus). Third-party drug names and regulatory agency names are referenced for identification only under the doctrine of nominative fair use.