Drug Repurposing Through AI-Driven In-Silico Clinical Trials

The pharmaceutical industry stands at a transformative crossroads, where traditional drug development paradigms are being revolutionized by artificial intelligence, electronic health records (EHRs), and the generation of synthetic data. Today's industry professionals are witnessing an unprecedented acceleration in drug repurposing methodologies that promise to reshape therapeutic discovery and development.

The current landscape of drug repurposing

Drug repurposing—the identification of new therapeutic uses for existing approved medications—has emerged as one of the most promising strategies to address the mounting challenges of traditional drug development. While conventional de novo drug discovery requires 10-17 years and costs between $2-3 billion per approved therapy with only an 11% success rate from Phase I trials, repurposing can reduce development timelines to 3-12 years with average costs of $300 million and success rates approaching 30%.

Recent market analysis reveals the drug repurposing market was valued at $35.33 billion in 2024 and is projected to grow at a compound annual growth rate of 4.9% through 2032. This growth is driven by several converging factors: cost-effectiveness pressures on pharmaceutical companies, the expansion of therapeutic applications through artificial intelligence (AI)-powered discovery platforms, and supportive regulatory frameworks that recognize the potential of computational approaches.

AI and machine learning: The new frontier

The integration of AI and machine learning into drug repurpose has fundamentally transformed the field's capabilities. Modern AI-driven approaches, exemplified by frameworks such as DeepDrug, now utilize graph neural networks to model the complex interactions between genes, proteins, and drugs within heterogeneous biomedical networks. These sophisticated systems can identify novel drug-disease associations by learning high-level representations from integrated genomic, proteomic, and clinical datasets.

A recent comprehensive systematic review identified 202 articles and 48 clinical trials that used in silico methods, of which 76 were linked to drug development. Notably, while most applications currently focus on diseases and imaging research, the methodology shows promise for rare and pediatric diseases that have historically been underserved due to limited commercial potential.

The DeepDrug methodology, for instance, recently identified a five-drug combination (Tofacitinib, Niraparib, Diltiazem, Pantoprazole, and Pravastatin) for Alzheimer's disease treatment by targeting neuroinflammation, mitochondrial dysfunction, and cholesterol metabolism pathways. This exemplifies how modern AI approaches can simultaneously evaluate multiple therapeutic targets and predict synergistic effects.

Electronic health records as clinical trial platforms

The utilization of EHRs for clinical research has evolved dramatically since early implementations. Current EHR-supported trials demonstrate remarkable cost-effectiveness, with per-patient costs ranging from $44 to $2,000—significantly lower than those of traditional clinical trials. The COVID-19 pandemic accelerated the adoption of integrated EHR-based research platforms, with quaternary healthcare systems successfully conducting large-scale clinical research projects entirely within commercial EHR systems.

The methodological framework established by researchers using the UK's Clinical Practice Research Datalink, encompassing over 11.3 million patients, demonstrates the power of population-scale EHR analysis. By implementing sophisticated propensity score methodologies and survival analysis techniques, researchers can emulate randomized controlled trials with unprecedented scale and generalizability.

Contemporary approaches have expanded beyond traditional high-dimensional propensity score methods to incorporate natural language processing and deep sequence models for estimating treatment effects. These advances address the challenge of extracting meaningful signals from the vast, unstructured data within EHRs, including physician notes and patient narratives that contain crucial contextual information.

Synthetic data and virtual patients

The emergence of synthetic patient generation represents perhaps the most revolutionary development in clinical trial methodology. Synthetic data can now augment or replace control groups in clinical trials, addressing ethical dilemmas around placebo administration and dramatically reducing patient recruitment requirements.

Recent implementations have demonstrated the practical utility of synthetic patients across diverse therapeutic areas. A leading European biotech company successfully leveraged synthetic data to develop CAR-T cell therapy, creating high-fidelity datasets from over 3,000 patients with non-Hodgkin lymphoma, acute lymphocytic leukemia, and solid tumors. This approach enabled confirmation of safety hypotheses and optimization of dosing regimens without exposing additional patients to experimental risks.

The FDA's recognition of synthetic data's potential is evidenced by their collaboration with industry partners on the ENRICHMENT In-Silico Trial Program, culminating in official guidance documents for computational modeling and simulation approaches. Regulatory agencies are increasingly open to synthetic control arms, though they emphasize the need for robust validation against real-world outcomes.

Contemporary success stories in anti-diabetic drug repurposing

The therapeutic class that originally inspired systematic EHR-based repurposing research—anti-diabetic medications—continues to yield remarkable discoveries. Recent meta-analyses confirm metformin's protective effects across multiple cancer types, with a comprehensive 2024 study of 166 publications demonstrating reduced risks for gastrointestinal (RR = 0.79), urologic (RR = 0.88), and hematologic cancers (RR = 0.87).

However, the repurposing landscape has expanded dramatically beyond metformin. GLP-1 receptor agonists, initially developed for diabetes management, now demonstrate efficacy across an expanding range of conditions:

• Neurological applications: GLP-1 agonists show neuroprotective properties in Parkinson's disease, with exenatide and lixisenatide reporting encouraging Phase II results and advancing to late-stage development. The mechanisms involve activation of neuronal survival pathways and reduction of neuroinflammation.
• Addiction medicine: Semaglutide and liraglutide demonstrate remarkable potential in alcohol use disorder treatment, with Swedish population-based studies revealing markedly reduced alcohol-related hospitalizations during treatment periods. This discovery exemplifies the power of real-world evidence generation through large-scale observational studies.
• Cardiovascular and renal protection: SGLT-2 inhibitors have achieved remarkable therapeutic expansion, with current approvals spanning heart failure with preserved and reduced ejection fraction, chronic kidney disease, and cardiovascular risk reduction. Recent evidence suggests additional benefits in the management of COPD, including decreased mortality.

Methodological innovations and technical advances

Modern drug repurposing research has evolved far beyond simple observational studies to incorporate sophisticated causal inference methodologies. The integration of inverse probability weighting with deep sequence models enables more robust estimation of treatment effects from longitudinal EHR data. These approaches address temporal confounding and selection bias that limit earlier observational studies.

The systematic approach pioneered in comprehensive analyses—encompassing 640 model specifications across multiple cancer types, treatment comparisons, and missing data handling strategies—has become the gold standard for robust signal detection in EHR-based research. This methodological rigor ensures reproducibility and minimizes false discovery rates that plagued earlier repurposing studies.

Contemporary implementations use advanced techniques to oversee missing data, including multiple imputations via chained equations and inverse probability weighting, to extract the maximum information from incomplete clinical records. These methods are crucial for EHR-based research, where data-missingness patterns often contain meaningful clinical information.

Regulatory evolution and industry acceptance

The regulatory landscape for computational approaches in drug development continues to evolve rapidly. The FDA's 21st Century Cures Act explicitly supports modeling and simulation approaches, while European regulators actively engage with synthetic data applications. Recent guidance documents provide frameworks for incorporating in silico evidence into regulatory submissions, with successful examples including Pfizer's use of computational pharmacology to bridge efficacy across different formulations without requiring additional Phase III trials.

Industry adoption has accelerated significantly, with major pharmaceutical companies establishing dedicated computational pharmacology and quantitative systems pharmacology groups. Companies like Roche, AstraZeneca, and BMS are successfully deploying AI-based models throughout the drug development lifecycle, from early asset prioritization to post-market surveillance.

Challenges and future directions

Despite remarkable progress, significant challenges remain in translating computational discoveries to clinical practice. Reproducibility concerns persist only 24% of published in-silico clinical trial articles provide open-source implementations, and only 20% make synthetic data publicly available. This limits independent validation and slows methodological advancement.

Data quality and interoperability challenges continue to constrain EHR-based research. The heterogeneity of clinical data systems, varying data quality standards, and complex privacy regulations create barriers to large-scale implementation. However, initiatives like the Canada Health Info Way and similar international efforts are addressing these infrastructure challenges.

The intellectual property landscape for repurposed drugs presents ongoing challenges, as patent protection for new indications of existing compounds remains complex. This can limit commercial incentives for comprehensive repurposing research, particularly for off-patent medications.

The path forward: Integration and innovation

The future of drug repurposing lies in the seamless integration of multiple technological advances: AI-driven target identification, EHR-based real-world evidence generation, synthetic patient modeling, and regulatory science innovation. Successful implementation requires collaboration across traditionally siloed domains, including academic research institutions, healthcare systems, pharmaceutical companies, and regulatory agencies.

Emerging applications in precision medicine show promise, where AI-driven patient stratification enables the identification of specific population subgroups most likely to benefit from repurposed therapies. This approach maximizes therapeutic value while minimizing development risks and costs.

The democratization of research capabilities through cloud-based computational platforms and standardized analytical frameworks promises to expand repurposing research beyond large pharmaceutical companies to academic institutions, biotech startups, and even patient advocacy organizations.

Conclusion

The transformation of drug repurposing through AI-driven methodologies, EHR integration, and synthetic data generation represents a fundamental shift in pharmaceutical innovation. As demonstrated by recent successes across various therapeutic areas, including oncology and neurodegenerative diseases, these approaches offer unprecedented opportunities to identify novel therapeutic applications for existing drugs, while reducing development timelines, costs, and patient risks.

For industry professionals, the imperative is clear: embrace these technological advances while maintaining rigorous methodological standards. The convergence of big data analytics, artificial intelligence, and regulatory innovation has created an environment where systematic drug repurposing can address unmet medical needs more efficiently than ever before.

The next decade is likely to witness the full maturation of in-silico clinical trial methodologies, with synthetic patients becoming routine components of regulatory submissions, and AI-driven repurposing platforms identifying therapeutic opportunities across the entire pharmaceutical landscape. Organizations that invest in these capabilities today will be best positioned to lead tomorrow's therapeutic innovations.

As we move forward, the integration of human expertise with artificial intelligence, real-world evidence with synthetic data, and traditional clinical trial methodologies with computational approaches promises to unlock therapeutic potential that has remained hidden for decades. The systematic approach pioneered in foundational research now provides a roadmap for transforming pharmaceutical development in the digital age.

The revolution in drug repurposing is not merely technological—it represents a fundamental reimagining of how we discover, develop, and deliver new therapeutic options to patients in need. For an industry facing mounting pressure to reduce costs, accelerate timelines, and improve success rates, these methodological advances offer a path toward more efficient, effective, and ethical pharmaceutical innovation.

Partha Anbil is at the intersection of the Life Sciences industry and Management Consulting. He is currently SVP, Life Sciences, at Coforge Limited

Jayanthi Anbil has over 15 years of experience in the Life Sciences Industry. Until recently, Jayanthi was with ICON Plc as a Global Business Intelligence Manager

Disclaimer: The views expressed in the article are those of the authors and not of the organizations they represent.