Some of the best-known drugs in modern medicine were not what they were originally designed to be. Sildenafil was a failed angina drug. Minoxidil was a blood pressure pill. Thalidomide was a notorious disaster that came back as a leukemia and myeloma medicine four decades later. The pattern repeats often enough that drug repurposing — finding new uses for existing drugs — has become a serious discipline in its own right.
This guide tells the stories of three of the most dramatic repurposing successes, then unpacks how computational repurposing is starting to find the next ones. You can run repurposing searches in the Drug Discovery Lab.
Why repurposing works
Developing a brand-new drug from scratch typically takes 10 to 15 years and costs billions of dollars. Most candidates fail. Repurposing an existing drug short-circuits much of that process: the molecule already has known pharmacokinetics, an established safety profile, and approved manufacturing. Clinical trials in a new indication can usually skip phase 1 because human safety is already characterized.
Beyond pure economics, repurposing works because biology is connected. Many drug targets are involved in multiple pathways, the same protein family (kinases, GPCRs, nuclear receptors) shows up in many diseases, and a drug that hits one of those targets often turns out to do interesting things elsewhere. The trick is finding those connections — sometimes in clinical observation, sometimes in computational search.
1. Sildenafil: from angina to Viagra
In the early 1990s, Pfizer was developing a phosphodiesterase type 5 (PDE5) inhibitor as a treatment for angina pectoris, the chest pain caused by reduced blood flow to the heart. The idea was that blocking PDE5 would prolong the action of cyclic GMP, relaxing vascular smooth muscle and improving coronary blood flow. The compound was called UK-92,480.
The angina trials were unimpressive. The compound did relax smooth muscle, but the effect on coronary arteries was modest and the safety profile was unremarkable. Pfizer was about to shelve the program. Then, in the male trial participants, an unexpected side effect started showing up consistently in patient questionnaires: improved erections.
Pfizer's scientists realized that the same PDE5 mechanism — relaxing smooth muscle by prolonging cGMP signaling — was ideal for treating erectile dysfunction. The corpus cavernosum of the penis depends on cGMP-mediated smooth muscle relaxation to allow blood inflow during erection. UK-92,480 was a near-perfect drug for that purpose. New trials were run with ED as the primary endpoint, and the results were striking. Sildenafil was approved as Viagra in March 1998. It became one of the most commercially successful drugs in pharmaceutical history.
Sildenafil has since been repurposed again — to treat pulmonary arterial hypertension, where the same vasodilatory mechanism is therapeutically useful in pulmonary blood vessels. It is sold under a different brand name (Revatio) for that indication.
2. Minoxidil: from blood pressure pill to Rogaine
Minoxidil was approved in the United States in the late 1970s as an oral antihypertensive. It works by opening ATP-sensitive potassium channels in vascular smooth muscle, causing arterial dilation and lowering blood pressure. It was used for severe, refractory hypertension that had failed other agents.
Patients on oral Minoxidil consistently reported an unusual side effect: their hair grew. Not just on the head, but all over the body. Hypertrichosis was so common that it became a practical limitation on the oral drug — patients (especially women) often discontinued because of the cosmetic effect.
Upjohn (which had developed the drug) realized that the unwanted side effect was, for some people, the desired effect. They reformulated Minoxidil as a topical solution that could be applied directly to the scalp, minimizing systemic exposure and concentrating the hair-growth effect where you wanted it. Topical Minoxidil was approved by the FDA in 1988 for androgenetic alopecia in men, and later for women. It is sold over the counter as Rogaine and remains one of the few evidence-based treatments for pattern hair loss.
The exact mechanism by which Minoxidil grows hair is still not fully understood. The opening of potassium channels in the dermal papilla may prolong the anagen (growth) phase of the hair cycle, and there is evidence for vasodilation of scalp microvasculature. Whatever the mechanism, the clinical effect is real and reproducible.
3. Thalidomide: from disaster to multiple myeloma
Thalidomide is the most famous repurposing story in pharmacology, and the most ethically complicated. It was introduced in West Germany in 1957 by Chemie Grünenthal as a sedative, with marketing that emphasized safety and a complete lack of side effects. It was widely prescribed to pregnant women for morning sickness.
Within a few years, thousands of children had been born with severe limb malformations — phocomelia — and other birth defects. The connection to Thalidomide was made in 1961 by Australian obstetrician William McBride and German pediatrician Widukind Lenz. The drug was withdrawn. The disaster prompted overhauls of drug regulation worldwide, including the strengthening of FDA pre-market approval requirements in the United States (where Frances Kelsey had famously refused to approve Thalidomide).
For decades, Thalidomide was a synonym for the worst-case scenario in pharmaceutical regulation. But in the 1990s, researchers noticed that the drug had immunomodulatory effects, including anti-angiogenic and anti-inflammatory activity. Trials in erythema nodosum leprosum (a complication of leprosy) showed dramatic benefit, and the FDA approved Thalidomide for that indication in 1998. Trials in multiple myeloma followed, and Thalidomide was approved for newly diagnosed myeloma in 2006 — combined with strict pregnancy-prevention programs.
The molecular mechanism was a mystery for years. In 2010, Hiroshi Handa's lab in Tokyo identified cereblon (CRBN) as the direct target of Thalidomide. Cereblon is part of an E3 ubiquitin ligase complex. When Thalidomide binds CRBN, it changes the substrate specificity of the ligase, causing it to ubiquitinate and degrade transcription factors that the myeloma cells depend on (notably Ikaros and Aiolos).
That discovery launched a new field of medicinal chemistry — molecular glue degraders and PROTACs — that uses drugs to induce targeted protein degradation. Thalidomide's derivatives Lenalidomide (Revlimid) and Pomalidomide (Pomalyst) became major multiple myeloma medicines, both through the same cereblon mechanism.
Computational drug repurposing
Sildenafil, Minoxidil, and Thalidomide were all discovered by clinical observation — somebody noticed an unexpected effect and chased it down. That works, but it's slow and depends on luck. Computational repurposing tries to make the process more systematic by using data to predict which existing drugs might work in which diseases.
Connectivity Map and gene expression signatures
The Broad Institute's Connectivity Map approach compares the gene expression signature of a disease to the gene expression signatures induced by thousands of approved drugs in cell lines. If a drug reverses the disease signature, it is a candidate for repurposing in that disease. This was the approach behind several COVID-19 repurposing screens in 2020.
Embedding similarity
Modern approaches encode drugs as chemical embeddings (from molecular fingerprints, graph neural networks, or pretrained language models on SMILES) and proteins as sequence embeddings (from ESM, ProtT5, or similar models). Drugs that are close to known target ligands in embedding space, or targets that are close to known drug-bound proteins, are flagged as candidates for further investigation.
Structure-based docking
With AlphaFold and ESMFold producing high-quality structure predictions for almost any protein, structure-based virtual screens against new targets have become routine. You can take a library of approved drugs, dock them computationally against a new target structure, and triage the top hits for experimental validation.
Knowledge graphs
Repurposing knowledge graphs link drugs, targets, diseases, side effects, and pathways into a single graph and apply graph neural networks to predict missing edges. A predicted drug-disease edge that doesn't exist in the literature is a repurposing candidate. This was an active approach during the COVID-19 pandemic and has continued to mature since.
Run a repurposing query in the Drug Discovery Lab
The repurposing endpoint in the Drug Discovery Lab takes a target or molecule and returns candidate drugs ranked by computational similarity. Here is the API call:
curl -X POST https://scirouter.ai/v1/drug-discovery-lab/repurpose \
-H "Authorization: Bearer sk-sci-..." \
-H "Content-Type: application/json" \
-d '{
"target": "EGFR",
"max_results": 20
}'The response includes ranked candidates with similarity scores and rationale. You can also seed with a SMILES instead of a target to find drugs structurally similar to a known binder.
Workspaces for the three classics
Each of the drugs in this article has a workspace in the Drug Discovery Lab:
- Sildenafil — PDE5 inhibitor, approved as Viagra and Revatio
- Minoxidil — potassium channel opener, approved as Loniten and Rogaine
- Thalidomide — cereblon binder, approved for ENL and multiple myeloma
Bottom line
Drug repurposing is one of the most efficient ways to bring treatments to patients. Existing drugs already have safety data, manufacturing, and pharmacokinetics worked out. The history of pharmacology is full of drugs that turned out to do something more interesting than the indication they were designed for — Sildenafil's journey from angina to ED, Minoxidil's pivot from hypertension to hair loss, and Thalidomide's redemption from tragedy to cancer therapy. Modern computational approaches are starting to find the next examples systematically, by combining data on drug structures, gene expression, protein targets, and disease biology into a single predictive framework.