Experiments
Two protocols, costed, for anyone who could run them.
This is the page most likely to change something. Both experiments are designed to be falsifiable, and a clean negative result from either would be published as readily as a positive one.
Experiment 2: The editing proof
Does the designed guide actually correct the letter?
US dollars in its minimum viable form, using an engineered cell line rather than patient cells. 26 weeks to the answer, 40 with function measured. Consumables and outsourced services only, with salary excluded.
What it gates
The base-editing manuscript, which at present contains no edited base at all. Everything in it is a prediction.
The design
Put the variant into a standard cell line, deliver the editor and guide, then sequence a designed 18-site panel to see whether the intended letter changed and whether anything else did.
How it resolves
It fails fast and cheaply. A week-12 checkpoint costing about 4,000 dollars shows whether any editing happened at all, and if not the remaining money is never spent.
Any detectable editing at MSH6, a DNA repair gene where loss of function causes an inherited cancer predisposition, ends the design. That is not a threshold to negotiate.
What it needs
A molecular biology bench, and nothing else. No electrophysiologist and no stem cells are required, which is the main reason this is the easier favour to ask.
Experiment 1: The mechanism
Does the broken copy simply sit out, or does it sabotage the working one?
US dollars when the patient and corrected cell lines already exist, rising to 59,342 if they have to be made. 42 weeks. Consumables and outsourced services only, with salary excluded.
What it gates
Four therapeutic routes, and the correction threshold for every one of them.
The design
Three cell lines, not two. The patient's own cells, the same cells with the letter repaired, and a third line with one copy deliberately switched off. That third line is the point: it makes the 45.8 percent benchmark a measured value rather than a theoretical one.
How it resolves
The decisive comparison is the patient line against the switched-off line. If the broken copy behaves like a silent one, the two sit on top of each other. If it interferes, the patient line sits below.
If the switched-off line does not read near 45.8 percent, the scale is broken and neither comparison means anything. That is the internal control, and it is stated in advance.
What it needs
A patch clamp rig and an electrophysiologist for roughly 34 recording days. That time, not the reagents, is the real cost, and it is excluded from the figure above.
Which one to run first
They are not the same experiment and the trade is real. The editing proof is roughly a third of the cost, needs no electrophysiologist, and reaches an answer in half the time. The mechanism experiment is worth more because it gates more: it decides the correction threshold for every remaining route.
If a laboratory can run only one, the editing proof is the better first favour to ask, because it is small, self-contained and fails fast. If a laboratory already has a patch clamp rig and an electrophysiologist, the mechanism experiment is the better use of that capability and nobody else appears to be doing it. The two share a cell line, so running them together saves roughly 8,000 dollars.
The detail that would have cost a month
The mechanism experiment needs to count the broken protein and the working protein separately at the cell surface, and no antibody can tell them apart. The standard way to do this is to cut the protein into fragments with an enzyme and weigh them.
The default enzyme cuts after arginine. Position 104 is an arginine, so it is itself a cutting site, and the variant removes it. The working copy gives a fragment four building blocks long, far too short to identify. The variant gives one twenty-one long. The two cannot be compared as a ratio because the change destroys the very site the method depends on.
A different enzyme, which cuts only after lysine, leaves position 104 untouched. Both copies then give the same twenty-six-block fragment, differing by 28.04 in mass, which any modern instrument separates easily. That was checked computationally before anyone was asked to run it.
What I am asking for
Not that anyone takes the analysis on trust. Every number traces to a named source with an identifier, and the working is public. The ask is narrower than a collaboration:
The full protocols, including every line item, the power calculation and the decision rules, are yours on request. Write to ethan@brugada.net. A correction is as welcome as an offer to run something.