---

Version of record: 10.5281/zenodo.21799860, published 5 August 2026. That identifier is the citable address for this paper and it resolves at https://doi.org/10.5281/zenodo.21799860. It is a version identifier; Zenodo minted a second one that resolves to all versions, and the version identifier is the one to cite.

What this file is. This project's authoritative copy of the manuscript, SUBMIT_THESE/papers/PUBLISH_5_R104Q_RECLASSIFICATION.md, which is the file the deposited PDF was built from. Synced 7 August 2026 by scripts/sync-manuscripts.mjs, which copies the source byte for byte and prepends this note. Nothing in the manuscript below has been rewritten for the website.

Warning, and it points at the published record rather than at this page. The version deposited on 5 August 2026 states, in two places, that no SCN5A expert panel exists to arbitrate this classification. That is false. The ClinGen Sodium and Calcium Channel Arrhythmia Variant Curation Expert Panel has SCN5A as its scope, completed the first of ClinGen's four expert-panel steps in March 2025, lists 27 members and is chaired by Andrew Glazer PhD. It was corrected here on 6 August 2026, after the deposit. The true and narrower statement, which the deposited note's own registry evidence supported, is that no released criteria specification covers SCN5A, so there is no expert-panel rule set to apply.

How the error arose, because that is the instructive part. ClinGen's Criteria Specification Registry and Evidence Repository were queried on 26 July 2026 and both correctly returned nothing for SCN5A. Those two resources record released specifications and expert-panel-classified variants; neither records whether a panel has been formed, and a panel that has completed step 1 of four has produced neither. An absence of a specification was read as an absence of a panel. The deposited note even checked that the empty result was not a broken query, by confirming the same tools returned records for other genes — a control that tests whether the question was asked properly and cannot test whether the right question was asked.

A second false statement in the deposited version, found the day after the first and dated 7 August 2026. Its data availability statement said three kinds of derived table were deposited: the recomputed statistics, the paralogue alignment counts, and the baseline-recalibration figures. Version 1 of the archive held one of the three. NTD_PARALOG_CONSERVATION.csv, the paralogue alignment counts, was there; the recomputed statistics and the baseline-recalibration figures were not. This is the same defect class as paper 10's and one degree less severe, and it is the one kind of claim a reader cannot check except by downloading the archive and finding nothing there.

Both missing tables were regenerable and both were regenerated, rather than the statement being quietly narrowed to fit what was there. They are arithmetic over published summary values, and this note's own Methods already record that the recomputation was done by hand from reported summary statistics with no software version, so there was no script to lose. p5_regen_statistics.py, written 7 August 2026 from the method as this note states it, reproduces all 32 quantities the note prints, and P5_RECOMPUTED_STATISTICS.csv and P5_BASELINE_RECALIBRATION.csv are staged with it, with its verification output P5_VERIFICATION_OUTPUT.txt and a provenance note P5_REGENERATION_NOTE.md. Nothing has been uploaded, so at the moment of writing the archive still holds only the one table. Two limits are stated rather than left to be discovered: the para-SAME sweep across 5,559 ClinVar variants is not deposited and is not claimed to be, and P5_BASELINE_RECALIBRATION.csv withholds raw per-cell values for 45 of its 51 rows because O'Neill's Supplementary Table 1 is a third party's table this archive should not redistribute in full.

One correction in this copy is against an earlier version of this copy, not against the deposit, and it is recorded because the pattern repeated. The first rewrite of the abstract said the expert panel "is developing" a criteria specification, while the body of the note says in terms that active drafting is an inference from a completed step 1 and an uncompleted step 2 rather than something the source asserts. The abstract asserted as fact what the body labelled an inference, and the abstract is what most readers see. It now states only the published status. The original error was reading an absence of output as an absence of a body; the abstract then read a completed step 1 as active drafting. Both are the same move, treating what a source permits as what a source states.

No measurement, no criterion, no points total and no conclusion changes. The variant still scores Likely pathogenic on both scored routes and Uncertain significance is still unsupported. What changes is the note's posture, and it improves it: it is an argument submitted to a panel that exists, not a stand-in for an expert judgement that nobody was going to make.

If a figure on this page disagrees with the same figure at the identifier above, this page is the corrected one. The full divergence, and what a version-2 deposit would have to include, is recorded in SUBMIT_THESE/ZENODO_DIVERGENCE_20260806.md. (This paragraph used to open by asserting that no version 2 had been deposited and nothing had been uploaded. That was true when written on 6 August 2026 and is not a claim a generated page can keep true, because it would turn false the moment anything is deposited and nothing here would notice. The sentence is removed rather than updated: to find out what is deposited, resolve the identifier, which is the only source that cannot go stale.)

None of this is peer reviewed, and none of it has been through a wet lab. No cell has been edited and no current has been recorded for this variant by this project. Every therapeutic statement in the manuscript below is a prediction.

---

The Conflicting classification of SCN5A p.Arg104Gln rests on one outdated submission, not on divided evidence

Ethan Bradley

Independent researcher, no institutional affiliation

ORCID: 0009-0008-8925-7975

Corrected 6 August 2026, in one place and everywhere that place reached. Version 1 of this note, deposited at 10.5281/zenodo.21799861 on 5 August 2026, stated that no SCN5A expert panel exists to arbitrate this classification. That is false. One exists: the ClinGen Sodium and Calcium Channel Arrhythmia Variant Curation Expert Panel, whose scope is SCN5A, which completed the first of ClinGen's four expert-panel steps in March 2025. The true and narrower statement, which version 1's own evidence supported and which this version makes, is that no released criteria specification covers SCN5A, so there is no expert-panel rule set to apply. The full account of what was said, why it was wrong, and how the error arose is in "Correction, 6 August 2026" before the references. The record deposited on 5 August 2026 carries the uncorrected claim. A second correction, dated 7 August 2026, follows it: version 1's data availability statement named three kinds of deposited table and the archive held one. Both missing tables have been regenerated from published summary values and deposited, and every file is now named by filename. No measurement, no criterion, no points total and no conclusion changes; what changes is who this note is addressed to, and it becomes a more modest document as a result.

Abstract

The ClinVar record for SCN5A p.Arg104Gln (NM_000335.5:c.311G>A; VCV000067780.15) currently reads "Conflicting classifications of pathogenicity," while the adjacent substitution at the identical residue, p.Arg104Trp (c.310C>T; VCV000067778.18), is Pathogenic/Likely pathogenic with no conflicts. I assembled the public evidence behind both records; the conflict does not reflect a genuine split. Of seven submitters applying classification criteria to R104Q, six call it Pathogenic or Likely pathogenic; the sole dissent, an Uncertain significance submission from a population-screening cohort last evaluated 6 October 2023, predates two calibrated functional datasets published in 2024 and 2025 (PMID 38953211, PMID 41251004). In those datasets R104Q and R104W are statistically indistinguishable on dominant-negative current and on automated patch-clamp Z-score. The identical substitution at the same invariant arginine is independently Pathogenic in SCN1A and SCN2A. Reanalysis of a 2013 oocyte dataset (PMID 23805106) shows R104Q shifts inactivation and slows recovery while leaving activation untouched, a specific gating defect, not simple loss of expression. The founding family in the original 2001 report showed low penetrance, weakening but not erasing the segregation evidence. A Bayesian points evaluation of ACMG/AMP criteria places R104Q at Likely pathogenic on two independent routes, not at Uncertain significance. No released ClinGen criteria specification covers SCN5A; the ClinGen expert panel with the gene in scope completed step 1 of four in March 2025. This is an argument for review by that panel, not a completed reclassification, and I am not a diagnostic laboratory. I am a heterozygous carrier of this variant; the implications are addressed under competing interests below.

A key to the terms used here

• SCN5A is the gene for the heart's main sodium channel; Nav1.5 is the protein. p.Arg104Gln, or R104Q, means arginine at protein position 104 replaced by glutamine. c.311G>A describes the same change at the DNA level.
• ClinVar is the public database of variants and laboratory interpretations. A VCV accession identifies the record for a variant; an SCV accession identifies one laboratory's individual submission within it. This distinction matters here, because the argument is about one SCV.
• ClinVar classifications run Pathogenic, Likely pathogenic, Uncertain significance, Likely benign, Benign. Conflicting is not a separate finding but a label ClinVar applies automatically when submitters disagree, which is the point of this note.
• ACMG/AMP criteria are the standard rules for classifying variants, with codes such as PS3 for supporting functional evidence. ClinGen maintains gene-specific versions of those rules.
• Penetrance is the proportion of carriers who actually develop the condition. Low penetrance means many carriers stay well, which affects how a variant is ascertained but not how damaging it is.
• Brugada syndrome is an inherited arrhythmia condition associated with reduced cardiac sodium current.
• Patch clamp is the electrical measurement of current through ion channels in a cell. Automated patch clamp does it at scale under one calibrated protocol, which is what makes variants measured on it comparable to each other.
• Dominant-negative means the variant copy interferes with the working copy, so the loss exceeds what one broken copy of two would cause.
• Oocyte expression means the channel was tested in a frog egg cell rather than a mammalian cell, an older method whose absolute numbers do not transfer cleanly to human heart.
• Paralogues are related genes from duplication. SCN1A R101Q and SCN2A R102Q are the equivalent substitutions at the equivalent position in two sibling genes.
• VEP is Ensembl's Variant Effect Predictor, a computational tool. Z score expresses a result in standard deviations from a reference mean.

Why this record is worth a second look

SCN5A encodes Nav1.5, the principal cardiac sodium channel, and loss-of-function variants in it cause Brugada syndrome. A variant's ClinVar classification shapes what a patient and their relatives are told, and what cascade testing follows. R104Q sits at a residue where a second, chemically different substitution has already been resolved without controversy. That asymmetry is the whole reason for this note. My interest in the question is not abstract: I carry R104Q heterozygously and have a clinical diagnosis of Brugada syndrome, a fact disclosed in full under competing interests.

Methods

All ClinVar data (VCV000067780.15, VCV000067778.18, VCV000068528, VCV000422127, and their constituent SCV records) were retrieved on 26 July 2026 via NCBI E-utilities esearch and esummary, with per-submitter assertions parsed directly from the VCV XML rather than transcribed from a summary page. Computational variant scores (AlphaMissense, REVEL, PolyPhen-2, SIFT) were retrieved on the same date from the Ensembl VEP REST service (rest.ensembl.org) against transcript ENST00000423572; the VEP release version is not recorded in my working notes. ClinGen's Criteria Specification Registry (cspec.genome.network) and Evidence Repository (erepo.clinicalgenome.org) were queried on 26 July 2026 to establish whether a released SCN5A criteria specification exists, and both were re-queried on 6 August 2026. Those two resources record released specifications and expert-panel-classified variants. Neither records whether an expert panel has been formed, and in version 1 of this note I read them as if they did. ClinGen's affiliation directory (clinicalgenome.org/affiliation/) answers that separate question and was queried on 6 August 2026; it is the source for the panel described below. Literature searches used Europe PMC full-text search restricted to the BODY field, with the terms "R104Q" and "R104W" jointly, and separately combined with "reclassification" and "conflicting"; PubMed and PMC were used to retrieve the primary functional and clinical papers cited throughout. Statistical recomputation (two-sample z-tests and Welch t-tests on published means and standard errors) was done by hand from reported summary statistics; no statistical software version is recorded for this step, so I state that plainly rather than inventing one. ACMG/AMP evidence was combined using the Bayesian points implementation described by Tavtigian and colleagues in 2020 (thresholds: Pathogenic at 10 or more points, Likely pathogenic at 6 to 9, Uncertain at 0 to 5); the full citation is given at reference 9 below.

R104Q and R104W move together in every assay that has tested them side by side

Both variants replace the same arginine at codon 104 of Nav1.5, one changing c.311G>A to glutamine and the other c.310C>T to tryptophan. Supplementary Table 1 of O'Neill and colleagues (Genet Med 2022, PMID 35305865) measured both on the SyncroPatch 384PE platform, first alone and then co-expressed with a genomically integrated wild-type allele. Expressed alone, current is abolished for both: 0.4 ± 0.2% of wild type for R104Q (n=22) against 0.5 ± 0.2% for R104W (n=24), a difference of -0.1 percentage points (95% CI -0.65 to +0.45, p=0.72). Co-expressed with wild type, the two remain indistinguishable: 68.3 ± 6.1% for R104Q (n=34) against 69.6 ± 7.3% for R104W (n=43), a difference of -1.3 points (95% CI -19.9 to +17.3, p=0.89). Because the wild-type allele in this assay is introduced by Sleeping Beauty genomic integration rather than diluted by the variant allele, the paper's own no-interference baseline is 100%, not 50%; against that baseline R104Q falls short by 31.7 points (z=5.20) and R104W by 30.4 points (z=4.16), both past the study's dominant-negative threshold of 75%.

The equivalence here is tight enough to bound, not just fail to reject. For the heterozygous measurement, no difference larger than about 16.9 percentage points (90% CI) is consistent with the data, and a true difference of 26.7 points or more would have been detected at 80% power. For the homozygous measurement the corresponding bounds are ±0.57 and ±0.79% of wild type. The honest statement is that no difference is resolvable at this sample size, not that the variants are proven identical.

A second, largely independent dataset agrees. The Vandenberg laboratory at the Victor Chang Cardiac Research Institute submitted evidence-only functional records to ClinVar for both variants (SCV007294411 for R104Q and SCV007294674 for R104W, both created 11 January 2026), citing the multi-site validation of an automated patch-clamp assay (PMID 38953211) and its cohort-scale application (PMID 41251004). R104Q scores a Z of -6.07 and R104W a Z of -6.17 on a scale where -4 marks severe loss of function; the submitting laboratory recommended PS3_strong for both, identically.

Computational predictors place the two in the same bin as well. AlphaMissense scores 0.8685 for R104Q and 0.8831 for R104W, both inside the 0.792 to 0.905 band that the published SCN5A calibration maps to PP3_supporting. REVEL gives 0.967 and 0.950; PolyPhen-2 and SIFT call both damaging without a calibrated strength. None of this distinguishes the pair.

The identical substitution at the same invariant arginine is Pathogenic in two other sodium channel genes

The residue is conserved outside SCN5A. SCN1A carries p.Arg101Gln (c.302G>A), classified Pathogenic in ClinVar (VCV000068528). SCN2A carries p.Arg102Gln (c.305G>A), also Pathogenic (VCV000422127). All three genes share the same nucleotide substitution, G>A, at the aligned invariant arginine. In an alignment of the nine human Naᵥ paralogue sequences, the residue is invariant at all nine positions; that count is my own, and the sequence source is not further specified in my working notes, so I flag it as this analysis's own computation rather than a cited figure.

The relevant framework for treating paralogue evidence this way, para-SAME as the paralogous analogue of ACMG PS1, was formalised and validated on the sodium-channel family by Brünger and colleagues (Genome Biol 2025, PMID 40624578), with positive likelihood ratios reported up to 13.0. I applied the criterion across the ClinVar pathogenic and likely-pathogenic missense variants available for the nine paralogues (5,559 variants) against a small internal answer key of variants with independent functional evidence of pathogenicity or benignity: R104Q, R104W, Y87C and F93S each recovered a para-SAME match, while R34C, Q90K and V125L, treated as benign controls, recovered none. That separation was complete on this small check set in both directions, which is encouraging but is a check on my own application of the method, not an external validation of it.

This line of evidence has a property the functional assays lack: it does not depend on interpreting current traces, expression systems, or baselines. SCN1A p.Arg101Gln and SCN2A p.Arg102Gln were classified by submitters with no connection to the SCN5A functional data or to each other. Its limits are real. SCN1A causes epilepsy, SCN2A causes epilepsy and autism, and SCN5A causes arrhythmia; para-SAME argues that the residue is intolerant of this substitution across the family, not that the diseases are equivalent. The two paralogue classifications are themselves expert judgements built on separate evidence bases, not independent experiments, and no released SCN5A criteria specification yet states how, or whether, paralogue evidence should be weighed in this gene (see below). (Corrected 6 August 2026: this clause read "and no SCN5A expert panel exists to weigh any of this". A panel exists; what does not exist is its released rule set.) Para-SAME contributes one criterion to an ACMG assessment; it does not reclassify anything by itself.

A thirteen-year-old oocyte dataset shows a specific gating defect, not simply lost current

The only dedicated functional study of R104Q, Gütter and colleagues (Front Physiol 2013, PMID 23805106), recorded it in Xenopus oocytes and reported gating parameters that the newer automated patch-clamp datasets do not tabulate. Their Table 3 gives, for wild type against R104Q: mid-activation voltage of -33.5 ± 0.4 mV against -34.0 ± 0.8 mV, a difference of -0.5 mV that is not significant (z=0.56, p=0.58); mid-inactivation voltage of -71.2 ± 0.7 mV against -73.8 ± 0.5 mV, a difference of -2.6 mV that is significant (z=3.02, p=0.003); and recovery time constant of 3.8 ± 0.1 ms against 4.6 ± 0.1 ms, a difference of +0.8 ms that is strongly significant (z=5.66, p<0.001). These z-tests are my own recomputation from the published means and standard errors and agree with the paper's own significance calls.

Activation is untouched. Inactivation shifts hyperpolarising and recovery slows. Channels that reach the oocyte membrane, at 29 ± 2% of wild-type current, are functionally abnormal in a specific, not general, way. That is different information from the O'Neill current-density measurements and it strengthens the case that R104Q is a genuine loss-of-function allele rather than an artefact of one expression system. It comes with an important restriction: these are measurements of R104Q expressed alone, the homozygous-equivalent condition, not co-expressed with wild type. They show that a mutant channel reaching the membrane gates abnormally. They do not show that R104Q disturbs the gating of a co-expressed wild-type channel, which is what the dominant-negative mechanism in a heterozygous carrier would require.

The two expression systems also disagree sharply in magnitude: essentially zero current in HEK293 cells against 29% in oocytes. Two explanations are both consistent with the data as it stands, a single defect that happens to be more severe in mammalian cells, or two separate defects, one affecting trafficking or assembly and dominant in mammalian cells, and an intrinsic gating defect visible only in channels that do arrive at the oocyte membrane. Nothing in the public record distinguishes these, and I do not claim one over the other.

The founding family had low penetrance, and that cuts both ways

A systematic search of full texts turned up a fourth independent pathogenicity assessment of R104Q that neither names the gene nor the variant in its title: Campuzano and colleagues (Sci Rep 2015, PMID 25608792) score it in body text as "previously reported associated with the disease, CM014904 (3 points), and with a deleterious effect in some studies (3 points). Positive familial segregation was identified in less than 4 relatives (2 points) but no in vivo studies were performed. In vitro studies showed a functional effect (2 points), and in silico analysis revealed four databases with damaging prediction (2 points)," for a total of 12 points, which the paper states indicates pathogenicity.

The HGMD accession CM014904 in that quotation, decoded by the format convention that encodes a first-report year rather than accessed directly in the database, points to the original description of the variant, Levy-Nissenbaum and colleagues (Genet Test 2001, PMID 11960580). Reading that source directly rather than relying on the 2015 summary matters, because its abstract states plainly that "the families of these patients are characterized by a small number of symptomatic members" and that "low penetrance is probably the cause for the small number of symptomatic members in the two families positive for the SCN5A mutations." The 2015 note's "positive familial segregation in less than 4 relatives" describes weak segregation, and the 2001 source explains why: the founding family itself had low penetrance. Carrying R104Q is evidently not equivalent to certain disease, and PP1-type segregation evidence here should be weighted with that in mind rather than treated as strong support on its own. I have not read the primary relative-by-relative segregation counts, since the 2001 paper's full text was not accessible to me, and I am not asserting a specific number of affected relatives beyond what the 2015 paper's summary states.

This also offers a plausible, but unconfirmed, reading of the single dissenting ClinVar submission. That submission comes from the All of Us Research Program, a population-screening cohort rather than a clinical referral laboratory. A low-penetrance variant ascertained in an unselected population is exactly the setting where a population-screening submitter might reach a different conclusion than clinical laboratories seeing symptomatic probands, an ascertainment difference rather than a disagreement about the strength of the functional or paralogue evidence. I present this as a plausible inference, not an established fact, since the submitter's internal reasoning is not published.

The same 2001 paper also reports "a possible R34C polymorphism in two unrelated controls," an independent confirmation from the primary clinical literature that R34C, used above as a benign paralogue control, was indeed seen in unaffected individuals.

The dominant-negative assay's own no-interference controls are not at 100%

Every dominant-negative statement above compares heterozygous current to an assumed baseline of 100%, one wild-type allele producing the full reference current with a dead mutant allele adding nothing. That assumption is testable against the 51 variants in the O'Neill 2022 dataset that have both homozygous and heterozygous measurements. Four of them are complete loss of function alone (homozygous current below 10% of wild type) yet show no dominant-negative effect when co-expressed: W822X (homozygous 4.7%, heterozygous 134.2 ± 5.2%, n=164), G1661R (5.6%, 112.0 ± 9.4%, n=44), S1672Y (1.0%, 100.8 ± 8.7%, n=47), and R893C (8.2%, 76.8 ± 10.8%, n=52), with a mean of 106.0% and standard deviation of 23.9% across the four. W822X, a nonsense variant truncated early enough that it cannot plausibly assemble into a multimeric complex, sits at 134.2%, 6.6 standard errors above the paper's assumed null of 100%.

Restating R104Q's effect against these alternative baselines changes its apparent size. Against the paper's own convention of 100%, R104Q falls short by 31.7 percentage points. Against the median of the four non-interfering controls, 106.4%, the shortfall is 38.1 points. Against the truncation control W822X alone, 134.2%, it is 65.9 points. The direction of the effect does not change under any baseline, and the comparison between R104Q and R104W is unaffected, since both share whatever baseline applies. What changes is the size of the claim: R104Q's dominant-negative effect is real but its magnitude sits somewhere between roughly 32 and 66 percentage points, with the range dominated by uncertainty in the assay's own baseline rather than by measurement noise on R104Q itself. I checked whether the paper's 75% dominant-negative threshold needs revision in light of this: only 2 of the 51 paired variants fall in the 75 to 106% window, and both, S1672Y and R893C, are among the four variants used to derive the 106% baseline in the first place, so that check is close to circular and has negligible practical effect on the study's published classifications. I record it rather than dropping it because the reasoning looked promising before the check showed otherwise.

Has this argument already been made in print

Europe PMC full-text search on the BODY field for "R104Q" and "R104W" jointly returns exactly one publication, Gütter 2013, which characterised R104Q functionally but performed no co-expression experiment and no classification analysis, and predates all of the dominant-negative and automated patch-clamp data discussed above. Searches combining "R104Q" with "reclassification" or "conflicting" return nothing relevant. The main text of the 2025 Eur Heart J paper (PMID 41251004) does not mention residue 104. That paper's supplementary tables S1 through S9 are behind a subscription and I could not open them. It is possible that R104Q already appears in one of those tables with a proposed classification; if so, this note would be corroboration rather than a new observation, and that possibility should be checked by anyone with access before this argument is treated as novel. I flag this honestly as an open question I could not resolve, not as a settled negative.

(Added 6 August 2026, as a sibling of the correction described at the head of this paper. This section is built on the same kind of inference that produced the expert-panel error, and deserves the same suspicion. A full-text search that returns nothing bounds the corpus searched, not the world. Europe PMC's BODY index does not cover subscription supplementary files — exactly the material this section already concedes it could not open — nor conference abstracts, nor unindexed preprints, nor the internal working documents of a variant curation expert panel, which is now known to have been meeting on this gene since March 2025. The claim this section supports is "I found no published statement of this argument in the corpus I could search", and that is weaker than "this argument has not been made". I did not restate it in the stronger form here, but the failure mode is the same one, and a reader should hold this section to the weaker reading.)

What this adds up to under ACMG/AMP criteria

No released SCN5A criteria specification exists, so there is no expert-panel rule set to apply to any of this. There is, however, an expert panel — a distinction version 1 of this note failed to make, and it is the correction described at the head of the paper.

The ClinGen Criteria Specification Registry held 206 specifications across 189 genes and 65 Variant Curation Expert Panels as of 26 July 2026, and none covers SCN5A; the only cardiac-rhythm specification available is for KCNQ1, an RYR2 panel is not yet released, and the five sodium-channel specifications that do exist (SCN1A, SCN2A, SCN3A, SCN8A, SCN1B, all from the Epilepsy Sodium Channel VCEP) do not apply to a cardiac channel. The ClinGen Evidence Repository confirms the same absence while returning records for KCNQ1, MYH7 and CDH1, so this is a true gap rather than a query failure. (That control is sound and is retained. It establishes that the query worked; it cannot establish that the query was aimed at the right question, and in version 1 it was not — see the correction before the references.) Generic ACMG/AMP criteria apply, informed by the assay-specific calibration published alongside the functional data.

(Re-checked 6 August 2026 from the registry's own record set rather than from its rendered pages: 206 specifications, of which 122 are flagged released and 84 are not, across 190 distinct gene symbols and 65 expert panels. SCN5A appears in none of the 206, released or not. KCNQ1 is released at version 1.0 under the Potassium Channel Arrhythmia VCEP, an RYR2 specification is present but not released, and the five epilepsy sodium-channel specifications are released at versions 2.0 to 2.1. A query to the Evidence Repository for SCN5A on the same date returned an empty result set, meaning no expert-panel-classified SCN5A variant exists there. The gene count differs by one from the 26 July figure of 189 above; I cannot tell from a single later snapshot whether a gene was added in the interval or whether I miscounted on 26 July, and I record the discrepancy rather than reconciling it silently. Nothing about SCN5A's absence changed between the two dates.)

The panel that would write that specification exists, and version 1 of this note said it did not. ClinGen's affiliation directory lists the Sodium and Calcium Channel Arrhythmia Variant Curation Expert Panel (affiliation 50160), under the Cardiovascular Clinical Domain Working Group, whose stated focus is the curation and interpretation of variants in SCN5A. Read on 6 August 2026, its four-step status tracker shows step 1, Define Group, "Completed Mar. 2025", and steps 2, 3 and 4 — Develop Classification Rules, Pilot Rules, Expert Panel Approval — carrying no completion date. It lists 27 members, a chair, Andrew Glazer PhD, and a coordinator, Emma Owens BSc. Its own description records that members have concluded SCN5A-related arrhythmias may be best treated as a merged spectrum phenotype, tentatively "SCN5A-related cardiac rhythm disorders", and that the group is in conversation with the Hereditary Cardiovascular Disease GCEP about re-examining SCN5A's curated gene-disease relationship. What I can state is the published status; that the panel is actively drafting rules today is an inference from a completed step 1 and an uncompleted step 2, not something the page asserts.

Three consequences follow, and they matter more to this note than the correction itself.

First, the framework used here is provisional by construction. Generic ACMG/AMP scoring is what a curator can apply to SCN5A today, and it is what an approved specification will replace. A points total assembled under generic criteria has no standing against an expert-panel classification, and I do not claim any.

Second, the disease framing may not survive. Every strength assignment below is made in a Brugada-specific frame — the PS3 calibration of PMID 38953211 is a Brugada assay, and the PM2_supporting frequency threshold is from the SCN5A-Brugada scheme. If the panel adopts the merged "SCN5A-related cardiac rhythm disorders" phenotype it is discussing, those calibrations are derived against a different disease entity than the one being classified, and would need re-derivation before they could be applied unchanged. This is a reason to treat the points total below as conditional on a framing that is under active review, not as a stable result.

Third, and least comfortable for this note: the panel is not short of the evidence assembled here. The panel's chair shares a name with the A. M. Glazer who is an author on all three of the functional datasets this argument leans on — the dominant-negative screen (reference 1), the assay validation (reference 2) and the cohort-scale application (reference 3). I am matching a name and a research subject, not a disambiguated identifier, so I state it as a strong identification rather than a verified one. If it holds, the measurements are not new to the body best placed to act on them. What is left as this note's own contribution is narrower than the length of the note suggests: the SCV-level observation that R104Q's Conflicting label is produced entirely by one 2023 submission that predates the evidence it would need to weigh, the side-by-side equivalence bounding against R104W, the paralogue argument, the reanalysis of the 2013 oocyte gating data, and the finding that the assay's own no-interference controls do not sit at 100%. That is a case for review, addressed to a panel that already holds most of the underlying data, rather than a case that no one else is in a position to make.

Under the SVI framework for PS3/BS3 (Brnich and colleagues, Genome Med 2019, PMID 31892348), evidence strength is tied to a calibrated OddsPath, with thresholds of 2.1 for supporting, 4.3 for moderate, 18.7 for strong, and 350 for very strong. The SCN5A-Brugada automated patch-clamp assay was calibrated against 49 controls (25 benign, 24 pathogenic), reaching 96% sensitivity, 96% specificity, and an OddsPath of 24.0 in the pathogenic direction (PMID 38953211), above the strong threshold. R104Q's Z-score of -6.07 sits well inside the severe bin the assay defines, and neither R104 variant was among the calibration controls, so PS3_strong applies without circularity.

R104W is a different missense change at the same residue with a clean Pathogenic/Likely pathogenic classification, which would ordinarily support PM5. The concern is double-counting if R104W's status rested on the same assay: it does not, since six of R104W's seven classifying submissions were evaluated between 2020 and May 2024, before the assay's August 2024 publication, so its status rests on clinical evidence gathered independently. PM5 applies at moderate strength; a curator who prefers strict independence may withhold it without changing the final bin (below). PM2_supporting applies because the ClinVar record reports a gnomAD exomes allele frequency of 0.00000 for R104Q, below the filtering threshold of 0.00003 used in the SCN5A-Brugada scheme; the gnomAD version behind that figure is not recorded in my working notes. PP3_supporting applies on the AlphaMissense score described above, and should not be taken above supporting.

Several criteria that might seem applicable are not. PS4 is not met: R104Q appears in 3 Brugada cases in the Walsh cohort, below the threshold of 5 needed for PS4_supporting. PS2/PM6 do not apply, since no de novo occurrence is documented and, in my own case, the variant is inherited rather than de novo. PP1 is withheld, because the Campuzano and Levy-Nissenbaum sources describe segregation in fewer than four relatives without giving the primary counts, and because the low-penetrance character of the founding family weakens what that segregation evidence would otherwise support. PVS1 does not apply, since this is a missense change regardless of how completely it abolishes current. PM1 is withheld, since the published scheme's N-terminal hotspot bonus depends on frequency and ancestry conditions I cannot verify for this variant from the public record. None of BS3, BP4, or BS1 apply, since every line of evidence runs in the pathogenic direction.

Combining these with the Bayesian points implementation (Pathogenic at 10 or more, Likely pathogenic at 6 to 9, Uncertain at 0 to 5): with PM5 included, PS3_strong (4) plus PM5 (2) plus PM2_supporting (1) plus PP3_supporting (1) totals 8, Likely pathogenic. Without PM5, the total is 6, still Likely pathogenic. Neither route reaches Pathogenic without additional clinical evidence such as verified segregation or stronger case enrichment, and I have not attempted to fold the SCN1A/SCN2A paralogue evidence into this points total, since none of the sources I consulted carries a worked Bayesian score that includes it; adding a PS1-strength code to an already Likely-pathogenic total would plausibly move the variant further, but I report that as an implication rather than a number I can verify. On either scored route, Uncertain significance is not supported by the evidence I could assemble.

The current shape of the ClinVar record

As retrieved on 26 July 2026, both VCV records carry nine submissions from nine submitters, seven of which apply classification criteria in each case. For R104Q (VCV000067780.15): Pathogenic from Labcorp Genetics (SCV000291795.7, 2025-09-30), Women's Health and Genetics/LabCorp (SCV005380503.1, 2024-08-20), the Montreal Heart Institute laboratory (SCV006065867.1, 2024-02-05), and Skåne University Hospital Lund (SCV005198180.1, 2023-09-19); Likely pathogenic from Ambry Genetics (SCV006422169.1, 2025-08-15) and GeneDx (SCV000235319.12, 2025-03-04); and Uncertain significance from the All of Us Research Program (SCV004834939.1, 2023-10-06). That is four Pathogenic, two Likely pathogenic, one Uncertain. For R104W (VCV000067778.18): Pathogenic from Labcorp Genetics (SCV002292504.5, 2026-01-07), Victorian Clinical Genetics Services (SCV005400264.1, 2023-07-17), Women's Health and Genetics/LabCorp (SCV004020423.1, 2023-06-12), the University of Leipzig laboratory (SCV003925662.1, 2023-04-19), and Ambry Genetics (SCV000738139.5, 2020-09-22); Likely pathogenic from GeneDx (SCV000518408.5, 2024-05-24) and Fulgent Genetics (SCV002800405.1, 2021-07-13). Five Pathogenic, two Likely pathogenic, no conflicts.

Four laboratories have classified both variants. Labcorp Genetics calls both Pathogenic. Women's Health and Genetics/LabCorp calls both Pathogenic. GeneDx calls both Likely pathogenic. Ambry Genetics calls R104W Pathogenic and R104Q Likely pathogenic, a real difference of one step that should be stated rather than smoothed over, but it is a difference between two pathogenic categories, not between pathogenic and uncertain. No laboratory that has classified both variants places R104Q in the uncertain category. The aggregate discrepancy in ClinVar's summary label is produced entirely by the single All of Us submission. Remove it, and R104Q's composition, four Pathogenic and two Likely pathogenic, closely matches R104W's five Pathogenic and two Likely pathogenic, with no conflict in either case. The dissenting submission cites "ACMG Guidelines, 2015" and was last evaluated on 6 October 2023, before the assay validation published in August 2024 (PMID 38953211), before the cohort-scale application published in November 2025 (PMID 41251004), and before the R104Q functional record itself was deposited in ClinVar in January 2026. A submission predating the evidence it would need to weigh is the ordinary reason a classification becomes outdated, not evidence of an unresolved scientific disagreement.

What would change my mind about this argument

Several specific findings would weaken or overturn what I have argued here. If the Sodium and Calcium Channel Arrhythmia VCEP releases an SCN5A criteria specification, the scoring below must be redone under it, and any classification that panel issues supersedes this note entirely rather than competing with it. That specification could move the total in either direction: it may calibrate PS3 differently from the assay paper, admit or exclude paralogue evidence, set a different frequency threshold, or score against the merged "SCN5A-related cardiac rhythm disorders" phenotype the panel is discussing rather than against Brugada syndrome. If the Eur Heart J 2025 supplementary tables, once accessible, show that R104Q has already been evaluated and placed in a category inconsistent with the argument above, that evaluation should take precedence over this note. If an independent laboratory using a different platform than SyncroPatch 384PE, and without shared senior authorship with the O'Neill and Ma groups, measured R104Q and R104W and found a real difference between them, the central claim of functional equivalence would need to be withdrawn. If the primary relative-by-relative segregation data in the 2001 Levy-Nissenbaum paper, once read directly, showed segregation inconsistent with a low-penetrance interpretation, the reading given here would need revision. If the All of Us submitter's internal reasoning were disclosed and showed genuine disagreement with the functional or paralogue evidence rather than an ascertainment difference, the reframing offered above would not hold. If co-expression data in a mammalian system showed that R104Q's oocyte gating defect disappears when the variant is expressed alongside wild type, that would separate the mechanism described here from the dominant-negative phenotype actually seen in heterozygous carriers, which is the discriminating experiment this note cannot itself perform.

What this note does not claim

This is an argument that a specific ClinVar submission should be reviewed against evidence published after it was made. It is not a reclassification, and I have no authority to issue one. It is addressed to the bodies that do have that authority — the submitting laboratories, and the ClinGen Sodium and Calcium Channel Arrhythmia Variant Curation Expert Panel, which has SCN5A in scope and is the right body to weigh this. Version 1 of this note argued in the belief that no such panel existed. It does, so this note is a submission to a process that is already running, not a substitute for one that is not. I am not a clinical diagnostic laboratory, I have not examined any patient, and nothing in this note constitutes clinical advice or a diagnosis for any individual, including myself. Classification decisions for this or any variant belong to a qualified clinical genetics laboratory applying its own validated procedures, and any question about what a given classification means for a specific person's care belongs with that person's own treating clinicians and genetic counselors, not with this document.

Data availability

All primary data used here are public. ClinVar records: VCV000067780.15 (R104Q), VCV000067778.18 (R104W), VCV000068528 (SCN1A R101Q), VCV000422127 (SCN2A R102Q), and the constituent SCV records named throughout, retrieved via NCBI E-utilities. Functional data: O'Neill et al., Genet Med 2022, PMID 35305865, Supplementary Table 1 (also available as bioRxiv doi:10.1101/2021.09.22.461398, Table S1); Ma et al., Circ Genom Precis Med 2024, PMID 38953211; O'Neill et al., Eur Heart J 2025, PMID 41251004, doi:10.1093/eurheartj/ehaf874; Gütter et al., Front Physiol 2013, PMID 23805106, Table 3. Clinical and segregation data: Campuzano et al., Sci Rep 2015, PMID 25608792; Levy-Nissenbaum et al., Genet Test 2001, PMID 11960580. Computational scores: Ensembl VEP REST (rest.ensembl.org) on transcript ENST00000423572. Framework references: Brnich et al., Genome Med 2019, PMID 31892348; Brünger et al., Genome Biol 2025, PMID 40624578. Registry checks: ClinGen Criteria Specification Registry (cspec.genome.network) and Evidence Repository (erepo.clinicalgenome.org), both accessed 26 July 2026 and both re-accessed 6 August 2026; the 6 August registry check was made against the registry's own underlying record set (206 records) rather than its rendered table, and the Evidence Repository check was the query gene=SCN5A, which returned an empty variantInterpretations array. Expert-panel status: ClinGen affiliation 50160, Sodium and Calcium Channel Arrhythmia Variant Curation Expert Panel (clinicalgenome.org/affiliation/50160/), accessed 6 August 2026. All derived tables are deposited as a single archive with a permanent identifier. The identifier is recorded in DATA_DOI.txt alongside this manuscript and should be cited as the data source. Version 2 of that archive holds, for this note: P5_RECOMPUTED_STATISTICS.csv, the recomputed statistics — every comparison above with its published group means, standard errors and cell counts, its difference and standard error, z, two-sided p and 95% confidence interval, the Welch t and its degrees of freedom where cell counts are available, the 90% equivalence bound and the difference detectable at 80% power; P5_BASELINE_RECALIBRATION.csv, the baseline-recalibration figures — the four non-interfering controls with their mean, standard deviation and median, W822X's distance above the assumed null, R104Q's shortfall against each of the three baselines, and the 75-to-106% window count, over all 51 paired variants of O'Neill's Supplementary Table 1; and NTD_PARALOG_CONSERVATION.csv, the paralogue alignment counts, whose row for position 104 records identity at all nine paralogues. The script that produces the first two from the published summary values, p5_regen_statistics.py, its verification print-out P5_VERIFICATION_OUTPUT.txt, and a provenance note P5_REGENERATION_NOTE.md are deposited with them.

Three limits on that statement, stated here rather than left to be discovered. First, version 1 of the archive contained only NTD_PARALOG_CONSERVATION.csv; the other two tables were named by version 1 of this note and were not there, and the correction below says so. Second, the para-SAME sweep across 5,559 ClinVar pathogenic and likely-pathogenic missense variants in the nine Naᵥ paralogues is not deposited and is not claimed to be — it was never among the tables this statement named, and it has not been regenerated. Third, P5_BASELINE_RECALIBRATION.csv carries raw per-cell measurements only for the six variants printed in this note, and marks the other 45 rows raw_values_withheld while keeping their derived flags, because O'Neill's Supplementary Table 1 is a third party's table that this archive should not redistribute in full; every figure printed here remains checkable from the deposited flags and summary block, and the table itself is public at the cited PMID and bioRxiv DOI.

Archive versioning. The concept DOI 10.5281/zenodo.21799233 always resolves to the current version of the data archive and is the identifier to follow for access. The version current at the time of this revision is version 2, 10.5281/zenodo.21840036. Version DOIs cited elsewhere in this manuscript name the specific version read and are deliberately not rewritten.

Correction, 6 August 2026

One claim in version 1 of this note was false, and it was a claim about who is allowed to decide the question the note is about. Version 1 stated, in two places, that no SCN5A expert panel exists. The ClinGen Sodium and Calcium Channel Arrhythmia Variant Curation Expert Panel exists, is affiliated to the Cardiovascular Clinical Domain Working Group, has SCN5A as its scope, completed ClinGen's step 1 in March 2025, lists 27 members and is chaired by Andrew Glazer PhD. The record deposited at 10.5281/zenodo.21799861 on 5 August 2026 carries the false claim.

How the error arose, because the reasoning is the instructive part. On 26 July 2026 I queried ClinGen's Criteria Specification Registry and its Evidence Repository. Both answered correctly: no released criteria specification covers SCN5A, and no expert-panel-classified SCN5A variant exists. I then treated those two absences as an absence of a panel. They are not the same thing. A criteria specification is an output a panel produces at the end of a four-step process; a panel that has completed step 1 of four has, by definition, produced none, and is invisible in both resources I checked while being entirely visible in the affiliation directory I did not check.

The part worth recording is that I did guard against error, and guarded against the wrong one. Version 1 explicitly checked that the empty result was not a broken query, by confirming the same tools returned records for KCNQ1, MYH7 and CDH1, and wrote that this made the finding "a true gap rather than a query failure". That control was valid and the query was not broken. It tested whether the question had been asked properly. It could not test whether the right question had been asked, and the failure here was the second kind: the tools answered "is there a released rule set" and I read the answer as "is there a body". A negative result bounds the thing actually queried and nothing else, and a control that confirms the query works will confirm it just as convincingly when the query is aimed at the wrong target.

The same inference appears once more in this note and is now flagged rather than corrected, since the sentence itself was already written in the weaker form: the novelty section infers from an empty Europe PMC full-text search that this argument has not been made in print. That search bounds the corpus it indexes, which excludes the subscription supplementary tables the section already concedes it could not open, and excludes the working documents of a panel now known to have been meeting on this gene since March 2025. A paragraph recording this sits in that section.

The abstract was tightened by fifteen words to pay for the twenty-three the correction added, and the edits are listed here rather than absorbed silently. The submission kit checks abstracts against a 250-word limit. Version 1's abstract measured 242 words; the correction added a twenty-word sentence and the three words "by that panel", taking it to 265; it now measures 250, which is at the limit and not under it. Seven phrases were shortened without changing any claim: "and find the conflict does not reflect a genuine split in the evidence" became "; the conflict does not reflect a genuine split"; "seven submitters who have applied" became "seven submitters applying"; "in SCN1A and in SCN2A" became "in SCN1A and SCN2A"; "The founding family described in the original 2001 report" lost "described"; "and does not support Uncertain significance" became ", not at Uncertain significance"; "a specific gating defect rather than simple loss of expression" became "a specific gating defect, not simple loss of expression"; and "that fact and its implications are addressed under competing interests below" became "the implications are addressed under competing interests below". No number, criterion, PMID or claim was removed. Every one of the seven is a wording change, and the substance of each is stated at length in the body section the abstract sentence summarises. The disclosure sentence is the one to check if you think a trim went too far; the full disclosure is unchanged under competing interests, and the section "Why this record is worth a second look" still carries it in full.

What does not change. No measurement, no recomputed statistic, no criterion assignment, no points total, and neither scored route. The conclusion remains Likely pathogenic on both routes, and Uncertain significance remains unsupported. Generic ACMG/AMP criteria remain the correct framework to apply today, because the thing that would replace them — a released specification — genuinely does not exist, which was version 1's evidence and remains true.

What does change is the note's posture, and it improves it. Version 1 implied, without stating it, that nothing else was coming: no panel, no rule set, so a self-taught carrier's note had to stand in for an expert judgement. That was the wrong picture. A panel exists, its chair is an author on the three functional datasets this note leans on, and it is developing exactly the rule set that will decide this question properly. This note is therefore not a substitute for expert review; it is an argument submitted to it, and a narrower contribution than version 1 believed itself to be making. That is a weaker claim and a better one.

Correction, 7 August 2026

A second false statement, in a different part of this note and found the day after the first. Version 1's data availability statement said three kinds of derived table were deposited. The archive held one.

Version 1 ended its data availability statement: "They comprise the recomputed statistics, the paralogue alignment counts, and the baseline-recalibration figures." Version 1 of the data archive (10.5281/zenodo.21799234) contained NTD_PARALOG_CONSERVATION.csv, which is the paralogue alignment count table, and neither of the other two. The record deposited at 10.5281/zenodo.21799861 on 5 August 2026 carries that claim.

This is the same defect class as paper 10's and one degree less severe, and it is a statement about a file in another place, which is the one kind of claim a reader cannot check except by downloading the archive and finding nothing there.

Both missing tables were regenerable and both were regenerated, rather than the statement being quietly narrowed to fit what was there. They are arithmetic over published summary values — two-sample z-tests on means and standard errors from O'Neill 2022 Supplementary Table 1 (PMID 35305865, also bioRxiv doi:10.1101/2021.09.22.461398) and Gütter 2013 Table 3 (PMID 23805106). This note's own Methods record that the recomputation "was done by hand from reported summary statistics" with no software version, so there was no script to lose. p5_regen_statistics.py, written 7 August 2026 from the method as stated in the Methods above, reproduces all 32 quantities this note prints: both R104Q–R104W differences with their confidence intervals and p-values, both pairs of equivalence bounds, all three Gütter z-tests, both dominant-negative z-scores, the four non-interfering controls recovered by this note's own stated rule and no others, their mean of 106.0, standard deviation of 23.9 and median of 106.4, W822X at 6.6 standard errors above the assumed null, the three shortfalls of 31.7, 38.1 and 65.9 points, the 51 paired variants, and the finding that exactly 2 of them fall in the 75-to-106% window and that they are S1672Y and R893C.

No number, criterion, points total or conclusion in this note changes. What changes is that the data availability statement now names its files, says what is not deposited, and is true.

One thing the regeneration turned up, recorded because it will trip anyone recomputing by hand. The Gütter differences printed above use the convention variant minus wild type, consistently across all three parameters. Computing them the other way round gives every magnitude correctly and every sign inverted. The deposited table follows this note's convention and states it.

One discrepancy between the Methods above and what the printed digits support, which I cannot resolve. The Methods say "two-sample z-tests and Welch t-tests". Every confidence interval printed here is a ±1.96 half-width, which is the normal multiplier, and every p-value matches the normal form at the digits shown. I could not identify a printed quantity that requires the Welch form. The two forms differ only in the third or fourth decimal here (homozygous p 0.7237 against 0.7254; heterozygous 0.8913 against 0.8917), so the paper's own digits cannot settle which was used by hand in July. The deposited table carries both. The Methods sentence is therefore left as written and flagged rather than rewritten to match the z form, because rewriting it would assert something about what I did that I can no longer verify.

Why version 1's statement was never caught. presubmit_check.py check 5 compares files cited by a paper against the deposit, and extracts citations with a regex that matches only backticked filenames. Version 1 described its tables in prose and named none, so the checker had nothing to match and reported nothing wrong. Six of the eleven papers in this set are invisible to that check for the same reason. This version names every deposited file in backticks specifically so that the check has teeth on it.

Three further abstract trims, listed rather than absorbed. The abstract sentence about the expert panel was weakened (see below) at a cost of four words, and the abstract was already at the 250-word limit. Three phrases were shortened without changing any claim: "is independently classified Pathogenic" became "is independently Pathogenic"; "on two independent scoring routes" became "on two independent routes"; and "which weakens but does not erase the segregation evidence" became "weakening but not erasing the segregation evidence". No number, criterion or PMID was removed.

And the abstract was overstating the panel, in the direction of the error it was correcting. Version 2 of the abstract, as first written on 6 August 2026, said the expert panel "is developing one" — a released criteria specification — while the body of this note says, in terms, that the panel "is actively drafting rules today is an inference from a completed step 1 and an uncompleted step 2, not something the page asserts". The abstract asserted as fact what the body labelled an inference, and the abstract is what most readers see. It now states only the published status: that the panel with the gene in scope completed step 1 of four in March 2025. The body is unchanged on this point because the body was already right, and one phrase in the ACMG section, "the panel that will eventually write that specification", lost "eventually" for the same reason. This is worth naming as a pattern rather than a slip: the 6 August correction was about reading an absence of output as an absence of a body, and the abstract then read a completed step 1 as active drafting. Both are the same move — treating what a source permits as what a source states.

Competing interests

I am a heterozygous carrier of SCN5A p.Arg104Gln, the variant discussed in this note, and I have a clinical diagnosis of Brugada syndrome.

Use of AI tools

This work was carried out with AI coding and research assistants (Anthropic Claude, via Claude Code). That use is disclosed here rather than left to inference.

Analysis code. The great majority of the analysis code in this project -- parsers, genome scans, regeneration scripts and verification scripts -- was written by an AI assistant working to my specification. I set what each script had to compute, chose the thresholds and the decision rules, and checked the output against the claims it is used to support.

Manuscript text. The prose of this manuscript was drafted by an AI assistant. I directed the drafting and revised the result, and I am responsible for every claim it makes.

Scientific decisions. The questions asked, the thresholds set, what was allowed to count as a refutation, and what was published were mine.

Verification, which does not depend on any of the above. Where a claim in this manuscript is regenerable from deposited inputs, the script that regenerates it and that script's own output are in the data deposit. Reproduction does not require trusting any account of who wrote what.

What no AI system did. No AI system generated, altered or selected any experimental measurement; this project contains no wet-lab data of any kind. All primary literature cited was retrieved from PubMed, PMC and publisher sources. Every reference in this manuscript has been machine-resolved against its own record, including a check that each PMID's first author and year match the author and year printed beside it in the text.

References

1. O'Neill MJ, Muhammad A, Li B, Wada Y, Hall L, Solus JF, Short L, Roden DM, Glazer AM. Dominant negative effects of SCN5A missense variants. Genet Med. 2022;24:1238-1248. PMID 35305865.
2. Ma JG, O'Neill MJ, Richardson E, Thomson KL, et al., Glazer AM, Ng CA. Multisite validation of a functional assay to adjudicate SCN5A Brugada syndrome-associated variants. Circ Genom Precis Med. 2024;17:e004569. PMID 38953211.
3. O'Neill MJ, Ma JG, Aldridge JL, Solus JF, et al., Glazer AM, Ng CA. Automated patch clamp data improve variant classification and penetrance stratification for SCN5A-Brugada syndrome. Eur Heart J. 2025. PMID 41251004. doi:10.1093/eurheartj/ehaf874.
4. Gütter C, Benndorf K, Zimmer T. Characterization of N-terminally mutated cardiac Na+ channels associated with long QT syndrome 3 and Brugada syndrome. Front Physiol. 2013. PMID 23805106.
5. Campuzano O, Sarquella-Brugada G, Cesar S, et al. Determining the pathogenicity of genetic variants associated with cardiac channelopathies. Sci Rep. 2015. PMID 25608792.
6. Levy-Nissenbaum E, Eldar M, Wang Q, et al. Genet Test. 2001;5(4):331-334. PMID 11960580.
7. Brnich SE, Abou Tayoun AN, Couch FJ, et al. Recommendations for application of the functional evidence PS3/BS3 criterion using the ACMG/AMP sequence variant interpretation framework. Genome Med. 2019;12:3. PMID 31892348.
8. Brünger T, et al. Paralogue-based variant interpretation in ion channel genes. Genome Biol. 2025. PMID 40624578.
9. Tavtigian SV, Harrison SM, Boucher KM, Biesecker LG. Fitting a naturally scaled point system to the ACMG/AMP variant classification guidelines. Hum Mutat. 2020;41(10):1734-1737. PMID 32720330. DOI 10.1002/humu.24088. (Resolved 10 August 2026; version 2 of this note carried the author and year only.)
10. ClinGen. Sodium and Calcium Channel Arrhythmia Variant Curation Expert Panel, affiliation 50160, Cardiovascular Clinical Domain Working Group. https://clinicalgenome.org/affiliation/50160/ — accessed 6 August 2026.
11. ClinGen. Criteria Specification Registry, https://cspec.genome.network/ — accessed 26 July 2026 and 6 August 2026. ClinGen Evidence Repository, https://erepo.clinicalgenome.org/ — accessed 26 July 2026 and 6 August 2026.

Back to all ten papers