Drug Repositioning Through Computational Modeling

Developing a new drug from scratch typically takes over a decade and costs billions of dollars, with roughly 90% of candidates failing in clinical trials. Drug repositioning — finding new therapeutic uses for existing, approved medications — offers a faster, cheaper alternative. The logic is straightforward: reuse what is already known about a drug's effects, strategically target the disease-relevant pathways, improve outcomes over the standard of care, overcome treatment resistance, and improve patients' quality of life.

This article traces how computational modeling enables that logic — from autoimmune disease, to oncology drug repurposing, to integrating the knowledge of natural compounds to enhance cancer therapy.

Figure 1: The computational drug repositioning workflow — from inputs to strategic pathway targeting to disease-specific applications.

Why Drug Repositioning Matters

The advantage of repositioned drugs is clear: they already have established safety profiles, pharmacokinetic data, and manufacturing processes — they can move to clinical evaluation far more quickly, dramatically reducing both cost and timelines.

The challenge shifts to efficacy prediction: which existing drugs might work for a different disease, and why? Computational methods address this by systematically screening approved compounds against disease-specific molecular signatures, strategically targeting the pathways that drive a disease to identify candidates that can improve on the current standard of care.

The Approach: Building a Virtual Disease

The core method builds disease models from the bottom up. It starts at the gene level — cataloguing which genes are active, mutated, or differentially expressed in a disease state. These genes map onto signaling pathways, and the interactions between pathways reveal how cellular behavior emerges: proliferation, apoptosis, inflammatory response, immune evasion.

From individual cell-type behavior, the model builds upward to cell-type interactions — how immune cells, stromal cells, and disease cells communicate within a tissue microenvironment. Finally, the tissue-level model extends to the systemic level, capturing how local pathology manifests as clinical disease.

This bottom-up architecture is what makes the approach versatile. The same gene-to-pathway-to-cell-to-tissue framework can be quickly reprogrammed to mimic a different disease — by modifying which genes are perturbed, which cell types dominate, and which microenvironment signals are active. A model built for rheumatoid arthritis can be adapted for cancer by reconfiguring the same pathway machinery with different disease drivers.

The human body is remarkably versatile yet fragile — small perturbations in a signaling pathway can cascade into entirely different disease states. The model leverages available genomic, proteomic, and clinical data to find the best therapeutic fit: which existing drugs, at which targets, in which combinations, offer the greatest chance of restoring balance.

Autoimmune Disease: Rheumatoid Arthritis

The standard of care for rheumatoid arthritis often involves immunosuppressive agents with significant side effects and incomplete efficacy. Patients frequently face treatment resistance and diminished quality of life despite being on therapy.

Systems biology-based in silico modeling was applied to understand anti-inflammatory mechanisms — how compounds like soluble epoxide hydrolase inhibitors modulate inflammatory signaling ^[1]. This established the approach: model the inflammatory response, then identify how existing compounds can be repositioned more effectively.

This evolved into predictive software-based mathematical modeling for rheumatoid arthritis — a novel approach that used computational models to design oral combination therapies from repurposed drugs whose mechanisms converge on the inflammatory pathways driving disease ^[2]. These combinations were validated in murine collagen-induced arthritis models, showing profound impact on disease progression and overcoming resistance to single-agent therapies ^[3]. Parallel work on oral inflammatory responses demonstrated that bidirectional immune regulation could be computationally predicted ^{[4], [5]}, providing a path to repurpose compounds for chronic inflammatory conditions.

This work resulted in patent filings ^{[6], [7]} and proved the concept: computational modeling can reduce the time and cost of finding effective repositioned therapies while delivering combinations that outperform existing treatments.

Oncology: Cancer Combination Therapy

Cancer rarely responds durably to single-agent therapies. Tumor cells adapt, compensatory pathways activate, and resistance emerges. Through the Beat AML project, computational biology models predicted patient responses to BET inhibitors, CDK4/6 inhibitors, dasatinib, venetoclax, and IDH inhibitors — validated against actual patient-derived drug sensitivity data ^[8].

The approach predicted not just whether a drug works, but for whom and why, based on strategically targeted pathway analysis — matching patients to drugs most likely to overcome resistance to the standard of care. Computational models also accurately predicted multi-cell biomarker profiles in inflammation and cancer ^[9].

Rational combination design demonstrated this further: modeling how nelfinavir (an HIV drug), metformin (a diabetes drug), and rosuvastatin (a statin) converge on cancer cell survival pathways in tumors with PTEN/TP53 aberrations. Three drugs designed for completely different conditions could target cancer through entirely new mechanisms ^[10]. The logic remained the same: reuse known effects, strategically target pathways, design combinations that enhance efficacy beyond single agents.

Natural Compounds: Expanding the Toolkit

Many plant-derived molecules have well-characterized mechanisms but have rarely been rigorously evaluated alongside standard cancer therapeutics using quantitative pathway-targeted methods.

Computational modeling was applied to identify plant-derived compounds whose mechanisms complement or enhance standard cancer therapies for hematological malignancies ^[11]. The framework strategically assessed which natural compounds targeted pathways that synergize with conventional drugs — enhancing therapeutic effect, overcoming resistance, and reducing side-effect burden.

This opens the same repositioning logic to a vastly expanded pool of candidate compounds.

Precision Medicine: Patient-Specific Repositioning

The most impactful evolution has been the shift from population-level to patient-specific predictions. Rather than asking "What disease could this drug treat?", precision repositioning asks "Which available drugs work best for this specific patient?"

In the Beat AML project, patient-specific genomic data drove individualized drug response predictions: characterize the molecular profile, simulate tumor response to different drugs and combinations, and compare against ex vivo drug sensitivity results ^[8]. By matching the right drug to the right patient, unnecessary treatment exposure is reduced, resistance patterns are overcome, and outcomes improve.

The Unifying Logic

Across autoimmune disease, oncology, and natural compound integration — the underlying logic remains consistent:

• Improve against the standard of care. Can we do better than what patients currently receive? Computational models compare repositioned combinations against existing treatments.
• Overcome resistance. Diseases adapt to single agents. Rational combinations target multiple pathways to preempt resistance.
• Improve quality of life. Known safety profiles and optimized dosing reduce treatment burden. Computational optimization balances efficacy with toxicity.
• Reduce research spend and time. Established safety and manufacturing data compress timelines from a decade to years, at a fraction of the cost.
• Reuse known drug effects. Existing pharmacological data predicts how effects translate to new disease contexts.
• Strategically target pathways to enhance effect. Quantitative pathway analysis identifies which combinations produce the greatest synergistic benefit.

Challenges

• Model validation. Predictions must be validated experimentally — in cell lines, animal models, and clinical settings. The most impactful computational work is tightly integrated with experimental validation.
• Data quality. Model accuracy depends on the breadth of underlying biological data. Gaps remain for rare diseases and underrepresented populations.
• Regulatory and IP complexity. New combinations require novel trial designs, and IP protection for repositioned uses can be complex. Early IP strategy is critical.
• Clinical translation. Pharmacokinetics, drug-drug interactions, and patient compliance introduce variables beyond computational models. Close collaboration between computational scientists and clinicians is essential.

Looking Ahead

Large-scale genomic databases, machine learning for drug-target prediction, multi-omics integration, and real-world clinical data are expanding what computational models can achieve. The core principle endures: build rigorous models of disease, strategically target the pathways that matter, and bring effective, affordable therapies to patients faster.

Drug repositioning will not replace traditional drug development, but it is an increasingly important complement — one where computational biology plays a central role in improving outcomes, overcoming resistance, and making drug discovery more efficient.