Introduction
Repurposing was at one time a very popular strategy, with some of the earliest blockbusters being novel formulations (usually new oral sustained release versions) of existing products. Generous regulatory pathways (like the FDA 505(b)(2) pathway) made rapid development a realistic strategy. But downward payer pressures and rapid generic entry made this strategy unfeasible for most companies.
The COVID-19 pandemic, coupled with emerging AI tools, threw repurposing back into the spotlight. Within months of the outbreak, researchers and AI systems scoured existing drug libraries for candidates. Baricitinib, a rheumatoid arthritis drug, was identified by AI as a potential COVID-19 treatment and received Emergency Use Authorization. Remdesivir, originally developed for Ebola, became a first-line treatment. While not all repurposing attempts succeeded, the pandemic demonstrated both the promise and the pitfalls of the approach.
Historically, the most successful examples of repurposing (i.e., sildenafil and minoxidil) were initially identified via serendipity, usually based on clinical observations or even observations noted after the drug was on the market. Today, repurposing is no longer the domain of serendipity alone. Systematic computational approaches, vast real-world data repositories, and increasingly sophisticated AI models are transforming drug repositioning from an opportunistic sideline into a core strategic capability.
Estimates vary, but some suggest that repurposed drugs have an overall success rate from Phase I to approval of approximately 30%, compared to 10-11% for novel compounds.
To leverage this potential, companies require a clear framework for identifying repurposing opportunities. Part I introduces an updated framework suited for the AI era, while Part II expands on it from an AI perspective.
The Repurposing Matrix
We propose a simple framework for categorizing drug repurposing opportunities based on two dimensions:
- Indication: Is the drug being developed for the same therapeutic indication it was originally designed for (old), or a new one?
- Formulation: Is the drug being delivered in its original dosage form (old), or a new one? By extension, does the new formulation imply a new route of administration, thereby enabling a completely new indication(s)?
This yields four distinct quadrants, each with different strategic implications:
| Old Formulation | New Formulation | |
| New Indication | Quadrant 3: Classic Repurposing | Quadrant 4: Full Transformation |
| Old Indication | Quadrant 1: Starting Point / Baseline | Quadrant 2: Reformulation for Optimization |
Quadrant 1: Old Indication, Old Formulation — The Starting Point
This quadrant represents the baseline where a drug begins its lifecycle before any repurposing occurs. A candidate is developed for a specific indication via a specific formulation. It may succeed and reach the market, or it may fail at some point in development. While this isn’t “repurposing” per se, Quadrant 1 is critically important because it establishes the foundation upon which all repurposing efforts build. The data generated here (or not generated, as the case may be) becomes the asset that enables movement into other quadrants.
We believe companies should make the effort to systematically catalog and characterize their Quadrant 1 assets (both approved products and shelved development candidates) to identify the richest opportunities for migration into other quadrants. Too often, this inventory exists only as fragmented institutional memory: old clinical study reports, archived INDs, or half‑remembered “failed” programs that were abandoned for reasons unrelated to biology or safety.
Every major pharmaceutical company should have a dedicated team tasked with auditing the company’s “graveyard” of shelved assets. Their mandate would be to use Agentic AI tools to screen shelved molecules against current biological targets to identifytheoretical repositioning opportunities (e.g., “This failed diabetes drug hits a receptor now known to influence fibrosis”). If the new indication is “off strategy”, they should package the data and seek to out-license the asset to a biotech or specialty pharma company focused on that niche. This turns a write-off into revenue (upfront payments + royalties) and ensures that potentially life-saving therapies get a second shot at reaching patients.
Quadrant 2: Old Indication, New Formulation —Reformulation for Optimization
In this quadrant, a drug continues to target the same disease but is delivered in a new dosage form. The therapeutic objective remains unchanged; what changes is how the drug reaches its target. Reformulation can address multiple objectives:
- Improved patient compliance: Converting injections to oral formulations
- Faster onset of action: Parenteral vs. oral administration
- Extended duration: Depot formulations for chronic conditions
- New clinical settings: IV formulations for hospital use
- Lifecycle management: Creating differentiated products to extend patent protection
For example, nalbuphine was a parenteral opioid which was reformulated as an extended-release oral tablet for the management of chronic pain, thereby migrating from Quadrant 1 to Quadrant 2. The program generated substantial Phase II data across multiple clinical models of acute and chronic pain. However, as an opioid, there was very little interest at the time to license and develop a novel opioid. The candidate was subsequently repurposed (and hence moved into Quadrant 3 from 2) for the treatment of cough associated with idiopathic pulmonary fibrosis (IPF).
Newer examples can be found on our formularies, with Wegovy perhaps being the most recent, high-profile example.
Examples
| Drug | Original Formulation | New Formulation | Benefit |
| Semaglutide (Wegovy/Rybelsus) | Injectable (weekly) | Oral (daily) | Patient convenience; needle-free |
| Risperidone (Risperdal Consta) | Oral (daily) | Long-acting injectable (biweekly) | Improved compliance in schizophrenia |
| Acetaminophen (Ofirmev) | Oral | Intravenous | Faster onset; inpatient surgical use |
| Ibuprofen (Caldolor) | Oral | Intravenous | Hospital setting; NPO patients |
| Paliperidone (Invega Sustenna/Trinza) | Oral | Monthly/quarterly injectable | Compliance; relapse prevention |
Conceptually, Quadrant 2 has the highest probability of technical success among the four quadrants, because the therapeutic hypothesis has already been validated. The drug works for the indication. The only question is whether the new formulation can deliver it appropriately.
However, regulatory and commercial risks remain. From a regulatory perspective, the new formulation must demonstrate bioequivalence or appropriate pharmacokinetic profiles. Novel delivery systems may require additional safety studies. From a commercial perspective, a marginal improvement in convenience may not support premium pricing and payer support.
Quadrant 3: New Indication, Old Formulation — Classic Repurposing
This is what most people think of when they hear “drug repurposing.” A drug that was developed (and possibly approved) for one indication is discovered to have therapeutic potential in an entirely different disease, without any change to its formulation.
Historically, many of the most famous repurposing successes arose from serendipitous observations:
- Sildenafil: Developed by Pfizer as an anti-anginal agent. During clinical trials, male participants reported a rather unexpected side effect. Pfizer pivoted, and sildenafil became the first oral treatment for erectile dysfunction, generating billions in revenue. Later, it was repurposed again (as Revatio) for pulmonary arterial hypertension, a drug that now exists in Quadrant 3 twice over.
- Thalidomide: Infamously withdrawn as a sedative after causing birth defects, thalidomide was later discovered to be effective for erythema nodosum leprosum (a complication of leprosy) and eventually for multiple myeloma. Its derivative, lenalidomide (Revlimid), became one of the best-selling cancer drugs in history, with $18 billion in sales in 2023.
- Minoxidil: Originally developed as an oral antihypertensive (Loniten), patients noticed increased hair growth. This led to the development of topical minoxidil (Rogaine) for male pattern baldness, although this example also crosses into Quadrant 4 because the formulation also changed.
- Aspirin: Originally marketed for pain and fever, aspirin’s antiplatelet effects were discovered decades later, leading to its widespread use for cardiovascular prevention. More recently, epidemiological studies have suggested a role in colorectal cancer prevention.
Some more modern companies have built their entire business model on “asset mining,” and even the largest pharma players are quietly adopting these tools. Here are a few examples of companies using AI to mine for repurposing opportunities:
1. In early 2020, BenevolentAI used its Knowledge Graph to scan thousands of approved drugs to find one that could both inhibit viral entry and reduce the cytokine storm. The company flagged baricitinib, a drug owned by Eli Lilly and already approved for Rheumatoid Arthritis. Because the safety profile was already known (Quadrant 3: New Indication, Old Formulation), it moved rapidly into trials. It showed a 38% reduction in mortality for hospitalized patients and received FDA Emergency Use Authorization in record time.
2. Lantern Pharma used their RADR® AI platform to analyze failed clinical trials and identify specific biomarkers that predict patient response. They identified LP-100 (Irofulven), a candidate originally abandoned after failing to meet efficacy targets. Lantern “mined” the data and realized the drug was highly effective, but only in patients whose tumors lacked a specific DNA repair mechanism. By identifying the right patients, they revived a “dead” asset and are now developing it for precision oncology.
3. Recursion takes a dual approach to asset mining. Instead of just mining literature, they screened a library of thousands of older molecules and found that a specific superoxide scavenger could treat Cerebral Cavernous Malformation (CCM), a rare genetic disorder. They successfully moved a repurposed asset into Phase 2 clinical trials, proving that AI can find “needles in the haystack” of existing chemical libraries.
4. Healx use AI to match the gene expression profiles of rare diseases against the known effects of existing drugs. The Asset: They identified a combination of therapies involving sulindac (an old anti-inflammatory drug) as a potential treatment for Fragile X syndrome. Eventually, they moved from prediction to Phase 2a clinical trials in a fraction of the time it takes for traditional development.
Platform Drugs: Deliberate Quadrant 3 Strategy
While serendipity played a role in early repurposing, modern pharmaceutical companies have learned to exploit Quadrant 3 systematically. Platform drugs, compounds with broad mechanisms of action, are deliberately developed across multiple indications. The biologics space provides some of the clearest examples:
| Drug | Original Indication | Subsequent Indications |
| Infliximab | Crohn’s disease | Rheumatoid arthritis, psoriasis, ankylosing spondylitis, ulcerative colitis |
| Adalimumab | Rheumatoid arthritis | Psoriasis, Crohn’s disease, ulcerative colitis, uveitis, hidradenitis suppurativa |
| Rituximab | Non-Hodgkin lymphoma | Rheumatoid arthritis, chronic lymphocytic leukemia, granulomatosis with polyangiitis, multiple sclerosis |
| Pembrolizumab | Melanoma | Lung cancer, bladder cancer, head and neck cancer, Hodgkin lymphoma, and 15+ other cancers |
These drugs didn’t just stumble into new indications. Their sponsors systematically tested them across diseases where the underlying mechanism (TNF-alpha inhibition, CD20 targeting, PD-1 blockade) was likely to be relevant. This represents the industrialization of Quadrant 3, driven by a mechanism of action and pharmacologic flexibility that makes these kinds of candidates incredibly valuable.
Quadrant 3 carries more uncertainty than Quadrant 2 because the therapeutic hypothesis must be re-validated in a new disease context. A drug that modulates a particular target may or may not be effective in a new indication, even if the biological rationale seems sound.
However, the advantages remain substantial via a known safety profile, established manufacturing, and an existing regulatory history.
Quadrant 4: New Indication, New Formulation—Full Transformation
This is the most ambitious quadrant: a drug is repositioned for an entirely new disease AND reformulated into a new dosage form. Both the therapeutic hypothesis and the delivery strategy must be reinvented. Sometimes, the new indication demands a different route of administration:
- Local vs. systemic effects: A drug with systemic side effects may be reformulated for local delivery to treat a localized condition
- CNS access: A drug targeting the brain may need reformulation to cross the blood-brain barrier
- Onset requirements: An acute indication may require faster onset than the original formulation provides
- Patient population differences: Pediatric or geriatric populations may require different formulations
Classic Examples
| Drug | Original Form/Indication | New Form/Indication | Rationale |
| Minoxidil | Oral / Hypertension (Loniten) | Topical / Male pattern baldness (Rogaine) | Avoid systemic hypotension; target hair follicles directly |
| Ketamine → Esketamine | IV / Anesthesia | Nasal spray / Treatment-resistant depression (Spravato) | Rapid CNS access; outpatient administration |
| Amphotericin B | IV / Systemic fungal infections | Liposomal IV (AmBisome) / Reduced toxicity formulation | Decreased nephrotoxicity while maintaining efficacy |
The Minoxidil Story: A Quadrant 4 Archetype
Minoxidil’s journey is perhaps the most instructive Quadrant 4 case study:
- Quadrant 1 (Starting Point): Minoxidil was developed as an oral antihypertensive. It was approved in 1979 as Loniten for severe, refractory hypertension.
- Serendipitous Observation: Patients on oral minoxidil reported unexpected hair growth (hypertrichosis) as a side effect.
- Quadrant 4 Recognition: Researchers realized that minoxidil might treat hair loss, but oral administration caused systemic hypotension, making it unsuitable for a cosmetic indication.
- Reformulation: A topical formulation was developed to deliver minoxidil directly to the scalp, minimizing systemic absorption.
- Outcome: Topical minoxidil (Rogaine) was approved in 1988 for androgenetic alopecia and became the first FDA-approved treatment for hair loss. It remains a market leader decades later.
In this case, reformulation was not optional. Without changing the delivery method, the new indication would have been impractical. Quadrant 4 opportunities often arise when the original formulation is incompatible with the requirements of the new indication.
Quadrant 4 is the highest-risk quadrant because it requires validating two hypotheses simultaneously: The drug must work in the new indication AND the new formulation must have the needed PK profile. However, the rewards can be substantial. Quadrant 4 transformations can create entirely new product categories and unlock markets that would otherwise be inaccessible.
Summary
Part I of this article proposes a simple framework for characterizing drug repurposing. We take a very broad view and definition of repurposing to include platform candidates, which is a strategy of increasing importance in our industry.
In Part II, we will take a look at just some of the many AI platforms and tools which are applicable across the entire repurposing framework. Today, AI can enhance repurposing by identifying new potential indications, optimizing formulations, and integrating candidate – formulation strategies… all at a speed which seemed impossible only a few years ago.