Limitations

What this work cannot tell you.

This is not a disclaimer page. It is a list of the specific boundary of each claim on this site, which is the part most easily lost when work gets summarised.

Nothing here was measured in a heart

Every functional number on this site comes from cells grown in a dish and engineered to produce this channel. That is the standard way to measure a variant, and it is not a heart. A heart has different cell types, different partner proteins, a different environment, and a gradient across the ventricular wall that no dish reproduces. This applies to the 68.3 percent figure, the 31.3 percent rescaling, and every statement about current.

Computational prediction nominates, it does not prove

The base editing guide, the off target scan and the variant nominations are predictions. A predicted off target site is a place worth checking, not a place where editing was observed. A nominated variant is a candidate for testing, not a variant known to be harmful. This project has direct evidence of how badly prediction can fail here: given four variants already known to break the channel, the folding stability method it had been relying on identified one.

What molecular dynamics simulations do and do not establish

The 1.41-million-atom all-atom simulation samples nanosecond timescales (up to 2.0 nanoseconds at 2 femtoseconds per step in a physiological POPC membrane with 120 mM NaCl). While nanosecond molecular dynamics captures local side-chain dynamics, contact excursions (such as the D84-Q104 distance reaching 7.33 Å), hydration column continuity (12.8 waters), and selectivity filter stability (DEKA radius 5.30 Å), it cannot directly simulate millisecond channel opening, macroscopic voltage-dependent inactivation kinetics, or hours-long trafficking through the Golgi apparatus. Those processes occur across microsecond to hour timescales and require physical electrophysiology and pulse-chase biochemistry.

Specific boundaries, by result

Off target scan for the base editing guide

Run across the genome, identifying 22 missense and 26 protein-changing candidate off-target sites with zero within two mismatches. Bulged alignments were modelled, but unexpected cleavage patterns in chromatin remain an experimental question.

The upregulation ceiling

Rules out one specific mechanism, boosting output by redirecting non productive message. It does not rule out every way of raising channel output.

Seventeen nominated variants

Candidates for testing. The other 114 are uninformative rather than benign, which is a different thing and matters to anyone carrying one.

The reclassification argument

An argument that the evidence should be reviewed. Only the submitting laboratories and the expert panel can change a classification.

The structural argument

One deposited structure, one conformation. Proteins move, and a single set of coordinates does not capture that. The 1.41-million-atom matched simulation samples dynamic motions and confirms pore hydration (5.30 Å DEKA filter, 12.8 waters), but operates on nanosecond timescales. Full gating state transitions require millisecond electrophysiology.

Where this project's own numbers disagree

Several quantities have been measured more than once here, by different methods, and do not agree. Where that is true no single value is quoted on this site. How buried position 104 is inside the assembled channel has five different values across five analyses. The count of protein changing off target sites for the lead editing guide was revised upward when a more thorough search was run, and the earlier figure still appears in places. The nearest protein changing off target moved closer once alignments with an inserted or deleted letter were modelled, so the earlier statement that it sat at three mismatches is no longer true.

What the experimental designs do and do not establish

The single channel measurement described on the experiments page has been derived and costed. It has not been run, and no part of it has been through a laboratory. Its central parameter was read from one published figure in one laboratory's hands, in a cell line, and is used to size an experiment rather than to make a claim. The whole derivation assumes that channels pair with each other at random, which is an assumption in this project, in the paper it argues against, and in the published literature generally. Nobody has measured it for any variant of this channel.

The route changes are readings of other people's published data, two of them digitised from figures rather than taken from tables. A value measured off a plotted figure supports about one decimal place, and both are quoted that way here. Where a paper's own statistics mark a comparison as not significant, that is stated rather than smoothed over, even where the comparison would help.

The correspondence answers specific technical questions and nothing broader. Where a researcher corrected this project, the correction is recorded on the page it affects. None of them has reviewed this work, and none should be read as endorsing it.

Not peer reviewed

None of these results has been through peer review. They are posted as preprints so the reasoning is visible early. That is a deliberate choice, and it means the appropriate level of confidence is provisional.