Integrated variant analysis · ion channels and accessory subunits

The channelopathy variant atlas

Select an ion channel gene from the catalog at left to fetch its complete ClinVar archive, gnomAD population variation, and AlphaMissense predictions in a unified, filterable view. Variants from each source are joined on protein change and rendered against the linear protein with annotated functional domains. A predictor at the bottom provides single-variant lookups for any residue substitution.

⟳ checking for AlphaMissense data…
Idle.
—
—
—
Length
—
—
ClinVar
—
missense + LoF
gnomAD v4
—
all variants
AlphaMissense
—
scored
ClinVar
AlphaMissense
Consequence
Range – gnomAD AF ≤ Search

Summary

— variants in current filters

Variant landscape

AlphaMissense Y · ClinVar color · domain track below
drag on plot to zoom · click marker to highlight in table

Structural mapping

AlphaFold predicted structure · variants mapped to residues
color by representation fragment click any residue to inspect · click a row or marker to spotlight
coverage: residues — — view on AlphaFold DB ↗
Click any residue in the 3D structure or any variant in the table to inspect.

Variants

— rows
rows per page ⌘/Ctrl+/ to focus search

Variant predictor

single-substitution lookup

Enter a missense substitution as one- or three-letter code. Numbering follows the canonical RefSeq/UniProt isoform shown above.

Results appear here.

Methods & data sources

ClinVar — variants are fetched live from NCBI E-utilities (esearch + esummary). The query is restricted to single-nucleotide variants and short indels with missense, nonsense, frameshift, splice, or synonymous molecular consequences. Each record's protein change is parsed and matched against the gene's canonical RefSeq protein numbering shown in the gene card.

gnomAD v4 — variants are fetched via the gnomAD GraphQL API at gnomad.broadinstitute.org/api using the gene.variants field on the canonical transcript (GRCh38). Reported allele frequencies are global; population-specific frequencies are available by clicking through to gnomAD.

AlphaMissense — pathogenicity scores are loaded from a local sidecar file (am_data.json) pre-extracted from the AlphaMissense_aa_substitutions release on Zenodo (DOI 10.5281/zenodo.10813168), filtered to the catalog's UniProt accessions. This means AM lookups are instant, work offline once the page is loaded, and cover every possible missense substitution in the canonical isoform — including ones absent from ClinVar and gnomAD. The build script is shipped alongside the HTML as build/build_am_data.py. Thresholds (from Cheng et al. 2023): < 0.34 likely benign, 0.34–0.564 ambiguous, > 0.564 likely pathogenic. If am_data.json is missing, the rest of the tool still functions but the AlphaMissense column will be empty.

AlphaFold structures — predicted models are pulled from EBI's AlphaFold DB API (alphafold.ebi.ac.uk/api/prediction). Proteins longer than ~1400 residues are split into overlapping fragments; the structure panel exposes a fragment selector. Rendering uses 3Dmol.js. Residue coloring modes: ClinVar classification (most severe variant per residue), AlphaMissense score, pLDDT confidence (AlphaFold's native scheme), or UniProt domain. Clicking any residue inspects its variants and synchronizes selection with the table and lollipop.

Domain annotations — pulled from UniProt's /uniprotkb/{accession}.json features (DOMAIN, REGION, MOTIF, TRANSMEM, BINDING types). Coordinates correspond to the canonical UniProt isoform; if you suspect register mismatch with your transcript, verify against UniProt's alignment.

Variant–variant matching — ClinVar and gnomAD records are joined on the parsed protein change (refAA·position·altAA). Records from one source without a match in the other are still displayed; the missing column is marked —.

Citation guidance

If a downstream report uses output from this tool, please cite the underlying primary sources: Landrum MJ et al., ClinVar 2024 (Nucleic Acids Res.); Chen S et al., gnomAD v4 (Nature 2024); Cheng J et al., AlphaMissense (Science 2023); Jumper J et al., AlphaFold (Nature 2021) and Varadi M et al., AlphaFold DB (Nucleic Acids Res. 2024); Rego N & Koes D, 3Dmol.js (Bioinformatics 2015); UniProt Consortium 2024.

Implementation: single-page application, no backend; all fetches occur in the user's browser. NCBI requests are throttled to ≤3/sec without an API key. The page caches gene data for the active session.