The Orphaned Pharmacopoeia
Medicines we already have, for diseases nobody tried them on.
- Category
- Biomedical research infrastructure
- Audience
- Researchers, funders, foundations, and clinicians
- Coverage
- 1,208 off-patent medicines
A ledger of medicine that already works, unused
The Orphaned Pharmacopoeia profiles every generic medicine approved in the United States — what it is made of, where in the body it reaches, what else its biology points at, and whether anyone has run the trial. The answer to that last question is usually no, for commercial rather than scientific reasons.
The stock is 1,208 off-patent medicines whose safety in humans is already established. 550 pairings of one medicine with one condition it is not approved for have been checked against trial records; 312 of those have no registered trial anywhere, and a further 28 reached Phase 2 — safety and dosing established — before being abandoned.
Most repurposing hypotheses fail at a step rarely checked before a trial is proposed: whether the drug reaches the target at a dose a person can tolerate. The site makes that the third thing on every profile, showing binding affinity against the concentration actually reached at the approved dose. Evidence is then re-weighted rather than taken at face value — genetic association carries 55% of the score and literature co-occurrence 2%, because co-occurrence saturates on any well-studied pair and measures publication volume rather than causation.
Every candidate is written as a hypothesis for falsification, with the reasoning shown link by link and a line naming what result would kill it. The closing sections answer the question the rest raise: no company can recover a trial on a molecule that has fallen to production cost, so the site prices the trial from first principles and names who might pay — NIH grants already funding that molecule, open federal programmes, and matched disease foundations.
Why proven medicines go untested
The incentive dies with the patent
Once a molecule is generic, the price has fallen to production cost. A company that funded a trial for a new indication could not recover the outlay — competitors could market the same molecule for that indication immediately, contributing nothing.
Reachability is assumed rather than checked
Most repurposing hypotheses fail on whether the drug engages the target at a tolerated dose, and that step is rarely tested before a trial is proposed. A target requiring higher concentrations than a patient can safely reach is not a candidate, whatever its in vitro potency.
Literature co-occurrence masquerades as evidence
A gene and a condition appearing together in published text saturates near a perfect score on any well-studied pair. Taken at face value it makes the most-written-about hypotheses look like the best-evidenced ones.
Phase 2 assets get abandoned and forgotten
28 of the pairings here had safety and dosing established and were then dropped. That work is already paid for and sitting unused, but there is no register that makes it findable.
The funding actually exists, in the wrong place
Trials of off-patent drugs are rarely funded by venture capital or pharma. They are paid for by government grants and disease foundations — and no one was connecting a specific molecule to the specific mechanisms open to fund it.
Provenance gets flattened in presentation
Measured binding data, derived pharmacokinetic estimates, and assumed cost placeholders end up rendered in the same typeface at the same confidence, which is how a placeholder becomes a citation.
Unanalysed reads as ruled out
A medicine missing from a repurposing list is easily taken as one that was considered and rejected. Without an explicit coverage statement, the gaps in a dataset become false negatives.
How we built it
Start from the whole generic formulary
Every generic approved in the United States — 1,208 medicines with human safety already established — rather than a curated shortlist. Brand-name molecules with no generic are searchable but kept out of the ledger, because a company still owns them.
Test reachability before biology
Distribution is predicted from molecular weight, lipophilicity, polar surface area and hydrogen-bond donors, and every binding target is judged in pAct against the plasma concentration reached at the approved dose. Targets that cannot be engaged in a patient are marked as such rather than listed as leads.
Re-weight the evidence instead of taking the headline score
Genetic association carries 55% of the evidence figure because it is the datatype most predictive of later clinical success. Literature co-occurrence carries 2%. Clinical and expression evidence are displayed in full but excluded from the ranking.
Write every candidate as a falsifiable claim
Each pairing shows the reasoning chain link by link, the evidence breakdown with its weights visible, any registered trial, and an explicit statement of what result would kill the hypothesis.
List the established uses alongside the candidates
Conditions the drug is already approved for are shown in their own section, so a reader can check the method against pairings whose answer is already known before trusting it on one that is not.
Score how abandoned each molecule is
A 0–100 figure built from manufacturer count, remaining exclusivity, and whether any reformulation route to exclusivity is still open — with 505(b)(2) named as the only remaining path on a generic, and the weights behind the score labelled unsourced.
Price the trial from first principles
Events needed follow the standard survival formula, enrolment divides through by event rate and follow-up, and the output is expressed as cost per person reached — the metric that matters to a funder with no revenue to set against the outlay.
Name who would actually pay
NIH RePORTER grants that already name the molecule, open Grants.gov notices with an explicit mandate to fund work on approved or off-patent molecules, and non-federal disease foundations matched to the body systems the drug touches.
Label provenance inline, every time
Measured, derived, and assumed appear next to the figures they qualify, so a placeholder cost-per-patient can never be mistaken for a quoted one.
State the coverage, not just the results
735 medicines carry measured binding targets and 110 have had candidate uses checked against trial records. The index says so plainly and states that the remainder are unanalysed rather than ruled out.
Capabilities
The full generic ledger
1,208 off-patent medicines, each with manufacturers, routes, patent status, and its own complete profile page.
Four views over the same stock
Every medicine, the best new uses, the patent line, and the 31 already used elsewhere — the same data cut by the question a reader arrived with.
Filter by what it could treat
Eighteen body systems, from brain and nerves through cancer, immune, metabolic and kidney, each carrying its own count.
Filter by what it is built from
Functional-group filters — grabs metal, sets where it goes, sticks permanently, sets how long it lasts, carries the main job — so a chemist can search by mechanism rather than by name.
Filter by how far anyone has taken it
Nobody has tried it, tried once then dropped, a big trial has run, or already standard. The first two are the whole point of the site.
Distribution profile
Where the molecule can actually get to — systemic circulation, central nervous system, intestinal lumen, skin, urinary tract — derived from its physical chemistry, because a drug can only act where it distributes.
Reachability-tested binding targets
Every measured target shown in pAct against the concentration reached at the approved dose, with the engagement threshold stated and unreachable targets marked as such.
Weighted evidence breakdown
Open Targets evidence types scored and re-weighted per pairing, with each type's contribution to the ranking shown next to its raw score.
Step-by-step reasoning chains
Each candidate use shows the inference link by link, with every link labelled by the provenance of the data behind it.
Falsification line on every candidate
A stated 'what would kill it' on each hypothesis — a null trial, or pharmacokinetic data showing the target is unreachable in the relevant tissue.
Abandonment scoring and the 505(b)(2) route
A 0–100 commercial-abandonment figure plus the one remaining route to exclusivity on a generic molecule: a new route or dosage form carrying three years of exclusivity.
Derived trial costing
An interactive model over effect size, event rate, follow-up and per-patient cost, producing events needed, enrolment, total cost, and cost per person reached.
Funder matching
NIH RePORTER grants naming the molecule, open federal programmes mandated to fund off-patent work, and disease foundations matched to the areas the drug touches.
Manufacturer register
Companies holding an approved abbreviated application from the FDA Orange Book, with approved products, routes, and the oldest product still on the market — potential partners, or objectors.
The visual language
The Orphaned Pharmacopoeia is built to read as a reference work under scrutiny. A grotesque display over a serif body gives long explanatory passages the weight of a monograph, while monospace labels carry the provenance tags — measured, derived, assumed — that qualify every figure. Colour is used only to encode body system and evidence tier, never for decoration.
Colors
Monograph Ink
#141B24
Primary type, headings, and data values
Paper
#F6F5F1
Base field, warm enough for long-form reading
Slate
#5F6E7B
Secondary copy, captions, and the second half of the headline
Deep Slate
#55636F
Table metadata and column labels
Signal Crimson
#E0245E
Never tested in people — the count the site exists for
Violet
#7B3FE4
New uses checked, and the off-patent status marker
Amber
#E08A00
Stopped after Phase 2 — work already paid for and abandoned
Verified Green
#00A86B
Established use and confirmed engagement states
Reference Blue
#2563E0
Generic medicine counts and cross-references
Rule
#E6E4DC
Table rules, card borders, and section dividers
Typography
Bricolage Grotesque
Sets the headline and section titles, with the qualifying half of each phrase dropped to slate. A wide grotesque gives the numbers presence without turning the page into a dashboard.
Source Serif 4
Carries the explanatory prose — the reachability argument, the evidence weighting note, the commercial rationale. A serif because these passages are meant to be read, not scanned.
IBM Plex Mono
Section eyebrows, filter group labels, identifiers, and the measured / derived / assumed tags. Monospace marks every place the site is declaring the status of a figure rather than asserting it.
Principles
- 01
Check reachability before biology
A target the drug cannot engage at a tolerated dose is not a lead. Putting that test third on every profile kills most hypotheses before they cost anyone a trial.
- 02
Weight evidence by what predicts success
Genetic association carries most of the score and literature co-occurrence almost none, because the latter measures how much has been written rather than what is true.
- 03
Every claim ships with its kill condition
Candidates are hypotheses for falsification. Stating what result would end each one is what separates a research instrument from a list of suggestions.
- 04
Provenance travels with the number
Measured, derived and assumed appear inline, so an unsourced placeholder can never be quoted back as a measurement.
- 05
Unanalysed is not ruled out
The coverage is stated plainly — 735 with binding data, 110 checked against trials — so absence from a result set is never mistaken for a negative finding.
- 06
Follow the money to the only funder left
Since no commercial actor can recover a trial on a generic, the site treats naming the grant, the programme, and the foundation as part of the finding rather than an afterthought.
Questions
A public ledger of 1,208 generic medicines approved in the United States, each profiled by what it is made of, where in the body it reaches, what its biology points at beyond its approved use, and whether anyone has ever run the trial.
Because the commercial case disappears with the patent. Once many manufacturers make a molecule the price falls to production cost, so a company funding a trial for a new indication could not recover it — a competitor could market the same molecule for that indication immediately without contributing.
A pairing has to survive three tests: the drug must distribute to where the condition is, it must engage the target at a concentration reached at the approved dose, and the target–condition link must hold up under evidence weighted toward genetic association rather than publication volume.
It is pAct — the negative logarithm of the molar concentration at which the drug half-occupies the target. Each whole step is a tenfold difference in potency. Oral drugs typically reach 0.1–10 micromolar in plasma, so roughly 6.0 is the threshold below which a target cannot be reached at a tolerated dose.
Literature co-occurrence measures how often a gene and a condition appear together in published text, which saturates near a perfect score on almost any well-studied pair. It reflects publication volume rather than causal evidence, so it carries 2% of the evidence figure against 55% for genetic association.
Yes. Each profile lists NIH RePORTER grants that already name the molecule, open federal programmes with an explicit mandate to fund work on approved or off-patent molecules, and non-federal disease foundations matched to the body systems the drug touches.
735 of the 1,208 medicines carry measured binding targets and 110 have had candidate uses checked against trial records. The rest have not been assessed yet. Absence from a result set means the work has not been done, not that the medicine was considered and rejected.
No. Candidate uses are unproven hypotheses derived from public data, and established uses are listed only so the method can be checked against what is already known. Nothing on the site is medical advice.
Interested in The Orphaned Pharmacopoeia?
Tell us how you would like to be involved — investing, building, advising, or simply following along — and we will be in touch.
Register Your Interest
Work with Vital Ventures
We study, support, and build the foundational systems that shape how societies function. If you are working on something in that direction, we would like to hear about it.
