T641K
Category 3/4 — Most DruggableConflictingTransmembrane · predictedEditorialThreonine → Lysine at position 641 inside wolframin's tenth transmembrane helix (TM10). ClinVar carries conflicting classifications — pathogenicity is documented but not universally confirmed. AlphaMissense 0.980, DynaMut2 ΔΔG -0.08 kcal/mol (essentially no destabilization). A near-zero-ΔΔG pathogenic variant with a striking mechanism.
Interactive 3D Structure
Bond changes · DynaMut2 interaction analysis
| Interaction type | Wild-type partner | Mutant partner | Status |
|---|---|---|---|
| Hydrogen bond | L637 | L637 | Preserved |
| Hydrogen bond | V638 | V638 | Preserved |
| Hydrogen bond | L645 | L645 | Preserved |
| Polar contact | L637 | L637 | Preserved |
| Polar contact | V638 | — | Lost |
| Polar contact | V644 | V644 | Preserved |
| Polar contact | L645 | L645 | Preserved |
| Van der Waals | L637 | L637 | Preserved |
| Van der Waals | L645 | — | Lost |
| Hydrophobic | F414 | F414 | Preserved |
Lost / gained / preserved interatomic contacts at the variant residue, from the DynaMut2 (Arpeggio) interaction analysis of the wild-type and energy-minimized mutant structures.
Computational Predictions
- transmembrane position; predictor benchmarked on soluble proteins
This variant sits in the transmembrane band. The stability value was computed with a predictor benchmarked on water-soluble proteins. Published benchmarks of ddG predictors on membrane proteins report correlation below 0.4. Treat the magnitude as unreliable and the sign as weak evidence only.
Do not rank this value against a variant from a different region on ΔΔG alone.
Clinical Evidence
Observed at very low frequency in gnomAD.
ClinVar classifies this variant as Conflicting classifications of pathogenicity for Wolfram syndrome (classic) (autosomal recessive, OMIM 222300); Wolfram-like syndrome (autosomal dominant, OMIM 614296). These are two clinically distinct disease families: classic Wolfram syndrome requires two affected copies, while the dominant WFS1-related disorders do not. One label covers both, and it does not tell you which applies to a given person.
- Wolfram syndrome (classic)autosomal recessive · OMIM 222300submitted as: Wolfram syndrome 1
- Wolfram-like syndromeautosomal dominant · OMIM 614296submitted as: Wolfram-like syndrome
- Submitters conflict on this classification (1★). Treat the conditions above as submitted, not as documented phenotypes.
- Phenotype scope is interpretation of submitted records. The computed values on this card (AlphaMissense, DynaMut2 ΔΔG, pLDDT) are measurements of the model, not of a patient.
- Presentation cannot be generalised from neighbouring variants: one nucleotide over can present very differently. Structural similarity supports a mechanistic hypothesis, never a phenotype claim.
Review status: 1★ conflicting classifications. ClinVar "conditions" is the union of all submissions; 0–1★ entries are submitted conditions, not documented phenotypes.
Highest in European (non-Finnish): AF 0.0013% (15 of 1,180,052 alleles), in line with the global figure.
No homozygotes reported in gnomAD v4. For a recessive allele this is expected and is not, on its own, evidence either way.
| Ancestry group | Allele freq. | AC / AN | Hom. | Carrier est. |
|---|---|---|---|---|
| European (non-Finnish) | 0.0013% | 15 / 1,180,052 | 0 | ~1 in 39340 |
Source: gnomAD r4, retrieved 2026-08-21. Values are allele frequencies. The carrier estimate is derived from the allele frequency (Hardy-Weinberg, ~2·AF) and is an approximation, not a published figure — where a published carrier frequency exists it outranks this column and is cited on the card.
Structural Context
Position 641 sits inside TM10, one of wolframin's eleven transmembrane helices anchoring the protein in the ER membrane. The AlphaFold model places T641 within 5 Å of immediate sequence neighbors LEU640 (2.5 Å) and ALA642 (2.5 Å), and into a hydrophobic cluster with LEU637 (3.8 Å), VAL638 (4.0 Å), PHE414 (4.1 Å, from TM3 — indicating helix-helix contact), and ILE643 (4.5 Å). The wild-type threonine's small polar character fits well in this membrane-embedded helix — its hydroxyl can participate in a localized hydrogen bond pattern within the helix backbone.
Replacing threonine with lysine in this position is striking. Lysine's positively-charged primary amine and its long alkyl chain are exceptionally costly in a transmembrane context: charge buried in the bilayer hydrophobic core is thermodynamically unfavorable, and the larger volume cannot be accommodated without local rearrangement. Yet DynaMut2 returns a |ΔΔG| of only 0.08 kcal/mol — essentially no destabilization.
The explanation is geometric. The lysine side chain, despite being large and charged, is flexible and can extend outward toward the membrane-water interface where the charge can be partially satisfied by interaction with lipid headgroups or water. The fold can absorb the substitution. But — and this is the mechanistic insight — the functional integrity of TM10 depends on its packing against TM3 (the PHE414 contact at 4.1 Å). Introducing a charge into that interface, even if the fold accommodates it, disrupts the helix-helix geometry that the wild-type relied on. The result is a variant the protein can fold but cannot function correctly.
AlphaMissense's score of 0.980 reflects this functional severity even though the structural cost is near-zero. The variant is pathogenic by mechanism, not by misfolding.
Druggability Assessment
The most likely mechanism is disrupted TM3-TM10 helix-helix packing: the introduced charge at the helix-helix interface (4.1 Å from PHE414 in TM3) is unfavorable in the bilayer hydrophobic core and would perturb the relative geometry of the two helices. The therapeutic strategy is site-specific: a small molecule that stabilizes the TM3-TM10 packing interface, occupying the geometric niche the wild-type threonine maintained.
This is a variant the Atlas captures particularly well — pre-atlas, the near-zero ΔΔG might have led screeners to deprioritize this position. The atlas tells you which interface to target.
Why this matters
Feed this card to Wolfram Intelligence
Download the T641K PDF below and upload it to Wolfram Intelligence to generate therapeutic-strategy proposals — guanidinium mimetics, sigma-1 agonist docking, NAC thiol-capping. NAC is already on the bench-testing list.