Methodology

The WFS1 Molecular Atlas is a complete, open, citable molecular atlas of Wolframin (UniProt O76024) — 2085 cards spanning every WFS1 variant class in ClinVar: missense, nonsense, frameshift, splice, in-frame indel, synonymous, and copy-number. Each card is a single, consistently-formatted record with the evidence appropriate to its type.

The 1152 curated missense cards integrate four independent evidence layers (AlphaFold, AlphaMissense, ClinVar, DynaMut2) with a wild-type/mutant 3D comparison. The remaining missense variants carry AlphaMissense + AlphaFold with DynaMut2 ΔΔG backfilling. Non-missense cards replace ΔΔG with the analysis that fits the mechanism: NMD status and retained-vs-lost domains for nonsense (52) and frameshift (78); exon-boundary and SpliceAI-pending calls for splice (154); sequence-level I1–I3 schema for in-frame indels (41); a silent call with splice-proximity flag for synonymous (517); and a gene-dosage view for copy-number variants (91).

Sources

• AlphaFold — structure model AF-O76024-F1 (v6) and per-residue pLDDT confidence, from the EBI AlphaFold Protein Structure Database.
• AlphaMissense — per-substitution pathogenicity score and class (likely benign / ambiguous / likely pathogenic).
• ClinVar — clinical classification, review status, associated conditions, and accession.
• DynaMut2 (Biosig, Univ. of Queensland) — predicted folding free-energy change (ΔΔG) on the AlphaFold structure.

Schema categories

• Category 2 — Moderately Destabilizing — 15 variants.
• Category 3/4 — Most Druggable — 348 variants.
• Category 4 — Stable Fold, Function Disrupted — 214 variants.
• Category 5 — IDR Exclusion — 143 variants.

Confidence and limitations

• IDR exclusion (pLDDT < 50). Variants in intrinsically disordered regions sit in low-confidence stretches of the AlphaFold model; their DynaMut2 ΔΔG is not clinically trustworthy and is flagged as Category 5.
• ΔΔG caveats. DynaMut2 predictions in transmembrane and lumenal contexts are less validated than for soluble globular proteins. The Atlas treats ΔΔG as a structural-tractability signal, not a clinical verdict. Experimental confirmation (DSF, thermal shift, CD) on a subset is the natural next step.
• AlphaMissense threshold. Scores above 0.564 are treated as the likely-pathogenic range; this is a calibrated threshold, not a hard biological boundary.

The headline finding

85 of 87 pathogenic & likely-pathogenic WFS1 missense variants (97.7%) land at |ΔΔG| < 2 kcal/mol — they do not gross-misfold (and 97.5% across all 608 missense variants with a stability prediction). Most pathogenic WFS1 missense variants damage the fold in measurable but tractable ways, implying the majority of carriers are candidates for small-molecule rescue rather than gene therapy. (ΔΔG is computed only where a DynaMut2 prediction exists; AlphaMissense-benign variants are intentionally left unscored.)

How to cite

Wallace M. WFS1 Molecular Atlas. RareResearch.AI, Free Life Enterprises, 2026. Per-variant citations are available on each card.

Foreword

A clinical foreword from Dr. Sarah Gladstone will appear here.

Contact & collaborate

RareResearch.AI · Free Life Enterprises. For data, validation collaborations, or clinical review, reach out via RareResearch.AI.