Two Doors in One Key

The two doors

The research paper names two access routes to the same cellular state: the voltage door and the redox door. The four classic agents sort cleanly between them. Family C fuses a voltage-seeking element and a redox-active element into a single molecule — three compounds deposited.

Why that is attractive

If the target state really is one condition reachable two ways, a molecule carrying both keys should reach it more reliably than either alone, and should be harder for a cell to adapt around.

Why it is the weakest claim in the set

Because it compounds two assumptions instead of testing one. A hybrid that underperforms both parents tells you little; a hybrid that works does not tell you which half did it. If any class in this series deserves scepticism first, it is this one, and the right experiment is the boring one: the hybrid against each parent, separately, same assay.

The deposited compounds

Addendum, 5 November 2026 — the branch was rebuilt

Nothing above has been altered. The original listing of four deposited redox-hybrid designs stands as published. What follows is appended.

The deposited lead of this branch, ARF-MOX-C4, was attacked on three grounds and did not survive any of them: the peroxide bond had to survive synthesis and plasma, the C5 ester could be cleaved by esterases before iron could activate the warhead, and the compound carried sixteen stereocentres of which six were unspecified — so it never defined a single synthetic target.

Then it turned out not to be the compound we had described. Ring perception on our own deposited structure gives a peroxide ring of C–O–O–C–C–C: six-membered, two ring oxygens, a 1,2-dioxane. An artemisinin-type 1,2,4-trioxane requires three. The appendage carries one carbonyl, so the linkage is a glutaryl-type monoester and not the succinate diester we reported. The structure governs and our prose was wrong. The correction cuts against the compound rather than for it: a plain dialkyl 1,2-dioxane has no stabilising acetal oxygen and is a poorer peroxide than the one we claimed, so the case for replacing it was stronger than the case made against it. The error never entered the structural record — every property ever published was computed from the correct structure — and it is corrected here by addition, not by deletion.

What improved, and what failed

The obvious repair, reading the antimalarial record, is the spiroadamantane ozonide of the OZ277/OZ439 lineage, where a cage shields the peroxide sterically. On this chassis it is a negative result: every adamantane design is worse than the compound it was meant to rescue on stereochemical definition, and no better on lipophilicity. We publish it as a failure because a reader deserves the dead end as much as the result.

What worked was symmetry. A tetraoxane with identical substitution either side of its peroxide ring has spiro carbons that are not stereocentres at all, so the ambiguity disappears with nothing having to be specified, while the molecule carries two peroxide bonds instead of one. A monospiro warhead bearing a methyl on the opposite ring carbon runs 50–100 Da lighter than the dispiro equivalent, and trading a spiro cyclohexane for an oxane exchanges hydrophobic surface for oxygen. Those moves reach clogP 6.39 where the first series bottomed out at 7.83.

Thirty-four designs, built independently on two planets and reconciled; twelve improve on the deposited lead simultaneously in lipophilicity, stereochemical definition and ester count. The deposited lead now ranks 31st of 34. S04 is the only design in the set to satisfy both Veber criteria with clean stereochemistry and is the lightest by 45 Da; S07 leads on the stated liability at clogP 6.60.

One thing is unresolved and it halves the branch. The C5 ester is the liability and the release mechanism: remove it and the conjugate can never revert to free moxidectin. A primary record has since been found — Naito and co-workers prepared C5-masked milbemycin ethers, esters, sulfonates and carbamates in 1994 and reported activity penalties across that set. That is a warning about retained parent activity, measured in a worm-motility assay rather than in cancer, and it is reported here against our own interest.

The thirty-four C4-branch designs, ranked best in class

In the PDF stack

• Chemical Cookbook of the Redox Series, Volume 6 — both rebuilt branches, every structure, and the synthetic routes.
• Rescuing an iron-activated peroxide conjugate — the C4 branch paper, 34 designs.
• Rescuing a mitochondriotropic conjugate — the voltage branch paper, 45 designs.
• ARF-DEF-002 defensive publication — patent-form disclosure, never to be filed, all rights waived.
• Corrigendum and taint register — which documents carried the misdescription and which did not.

Published designs, not medicines. Every property is computed. Nothing in either branch has been synthesised, assayed or administered, and no compound has biological data of any kind.