WE DESIGNED FOURTEEN CANCER DRUGS AND WE ARE GIVING THEM AWAY

OSDC Field Intelligence · Open Source Data Collective — A defensive publication. Everything below is released into the public domain as prior art on 2 November 2026.

We are not selling it, licensing it or holding it back. The point is that nobody else can own it either.

We spent one night designing fourteen cancer drug candidates that have never existed. This morning we are giving them away. Every structure, every formula, every machine-readable string is below — released as prior art so that no company can ever patent this chemistry, charge you for it, and tell you it works for reasons only they understand.

Why we are doing this

There is a hole in how medicine gets made, and it is not a conspiracy. It is an incentive. A drug nobody can own is a drug nobody will pay to test. Fenbendazole and ivermectin are cheap, off-patent and well understood, which is exactly why no sponsor will fund the large controlled trials that would settle what they can and cannot do.

The Repurposing Drugs in Oncology project exists because of this gap. It is a market failure, and it costs lives quietly.

The standard answer is to invent something new and patent it. We did invent something new. We are choosing the other door: publish it, so the chemistry belongs to everyone and to no one. From today these structures are prior art.

Any patent office in the world can cite this page against an application. We have given up our own right to own them by doing this, deliberately and with our eyes open.

If a laboratory wants to make these and test them, they are free to. We would like to know how it goes.

The reasoning, in order

1. Four unrelated drugs hit the same thing

Ivermectin, the benzimidazoles (fenbendazole, mebendazole), the HIV protease inhibitors and the artemisinins share no common molecular structure. They were designed for parasites, viruses and malaria. Yet across dozens of published studies they converge on the same outcome in cancer cells: disrupted mitochondria, altered membrane voltage, collapsed glucose handling.

In ordinary pharmacology that does not happen. Unrelated keys do not open the same lock.

Our reading is that the target is not a receptor at all. It is a state — the cell’s bioelectric and redox condition.

If that is right, these drugs are not keys. They are things that detune a state, which is why their chemistry can differ so wildly while their effect does not.

2. The problem with ivermectin

Ivermectin’s anticancer concentrations in the laboratory are far above the blood levels a safe human dose produces. That gap is the honest objection to the whole repurposing case, and no amount of enthusiasm closes it. A drug that works in a dish at a concentration you cannot reach in a person is a lead, not a treatment.

3. Why moxidectin is the better scaffold

Moxidectin is ivermectin’s neglected cousin, and on four independent measures it is the superior starting point:

• FDA-approved for humans in 2018 for river blindness — the first new drug for that disease in thirty years, with roughly 1,500 patients in the trials behind it. The safety package already exists.
• Terminal half-life of about 18–23 days, against roughly 18 hours for ivermectin. A slow effect on cell state needs sustained exposure, not a spike.
• More lipophilic (log P 5.4 versus 4.3), so it loads deeper into tissue.
• A poor P-glycoprotein substrate. This is the decisive one. P-gp is the pump tumour cells use to throw drugs back out. In P-gp-deficient mice, ivermectin exposure rose while moxidectin’s did not move at all — the escape hatch that caps ivermectin is largely closed to it.

And the published anticancer data is in a plausible range. Song and colleagues reported in Oncology Reports (2018) that moxidectin reduced glioma cell viability, arrested the cell cycle and shrank tumours in live animals.

A follow-up in the Journal of Cancer (2020) traced it to autophagic cell death through AKT/mTOR inhibition. Separate work reports antiproliferative activity at 25–100 nanomolar — a concentration blood can plausibly reach, unlike ivermectin’s micromolar requirement.

No human cancer trial of moxidectin has ever been run. Not one, anywhere, in the eight years since that paper.

4. The three doors

If the target is a state rather than a receptor, there are three ways in, and a molecule can be built to use each:

• The metabolic door. Tumours import glucose avidly through GLUT transporters. Hang a sugar on the drug and the tumour’s own appetite carries it inside, where an esterase cuts it loose.
• The voltage door. Attach a positively charged, fat-soluble group and it accumulates in mitochondria in proportion to membrane voltage — meaning the abnormal state is itself the delivery mechanism. The sicker the cell, the more drug it takes.
• The redox door. Attach a quinone or an endoperoxide and one molecule carries two orthogonal stresses with identical pharmacokinetics, which a mixture of two drugs can never do.

5. Where to attach it

Moxidectin has exactly two free hydroxyl groups: a tertiary one at C7 and a secondary, allylic one at C5. C5 is the classical, synthetically accessible handle of this chemical family — the same position modified in the known 5-oxime chemistry of the avermectins. Every compound below is a C5-O-acyl conjugate, leaving the spiroketal, the C23 methoxime and the macrocycle untouched.

The general form is MOX–O–C(=O)–L–T, where L is a linker and T is the targeting or pharmacophore group.

The full record: fourteen compounds

All structures below were built and valence-validated in RDKit 2026.09.1. Properties are computed, not measured. None of these compounds has been synthesised.

None has touched a cell. This is a design disclosure, not a result.

Molecular weight in daltons; clogP is computed lipophilicity; TPSA is topological polar surface area in Ų; HBD is hydrogen-bond donors; RotB is rotatable bonds. Note that family A drops clogP by up to 2.3 units and roughly doubles TPSA — these are designed to be transporter-dependent rather than passively diffusing, which is the inventive step over the parent.

Machine-readable structures (SMILES)

Copy these directly into RDKit, ChemDraw, PubChem or any cheminformatics tool.

The honest state of it

These fourteen compounds exist as designs. None has been synthesised; none has touched a cell. That is what a design disclosure is, and we are publishing it at exactly this stage on purpose — because the moment it is public, it cannot be enclosed.

Moxidectin itself is an approved antiparasitic with preclinical anticancer data and no human oncology trial. The conjugates are our chemistry, built on that scaffold, reasoned from the convergence.

One instruction, and we mean it plainly. Nobody should delay or drop cancer treatment over anything on this page. Patients who take alternatives instead of standard care die at substantially higher rates — that is measured, not asserted. This is chemistry for a laboratory to test, and for a patient it is a conversation to have alongside treatment, with an oncologist.

And moxidectin is not a thing to improvise with. Its half-life runs to three weeks, so an overdose cannot be undone — it accumulates while you wait. Veterinary formulations are dosed for animals many times human size.

What would settle it

Three experiments, none expensive, none run:

1. Make three of these and test them. Synthesis and a 60-cell-line panel with membrane-potential and mitochondrial readouts: roughly $80,000–250,000. If the conjugates do not beat the parent compound, the idea is wrong and we would like to know.
2. Does potency track cell state? Measure anticancer potency across the whole incoherent drug list against measured membrane and mitochondrial potential, controlling for how fast the cells divide. If it tracks molecular descriptors instead, the convergence thesis dies.
3. Pulsed versus continuous dosing at equal total exposure. Clinicians already pulse these protocols by instinct. Nobody has tested whether the instinct is right.

It is yours

If you run a laboratory, take it. No licence, no fee, no permission, no conversation with us required.

We have put it beyond ownership on purpose, which means it is also beyond our control — somebody may make these and find they do nothing at all. That is a result too, and a more useful one than another decade of nobody looking.

What we would ask, and cannot enforce: if you test them, publish what you find, including the nulls. That is the whole disease we are trying to treat here.