A31G

Position 31 sits in wolframin's N-terminal IDR. The pLDDT score of 28 indicates that AlphaFold cannot reliably predict the local conformation — the protein adopts an ensemble of structures in this region rather than a single fold. Neighbor analysis returns only the immediate sequence neighbors (SER32 at 2.5 Å, THR30 at 2.6 Å, LEU33 at 4.6 Å) — the structural signature of IDR.

Replacing alanine with glycine adds backbone flexibility. Glycine permits backbone conformations (especially in the Ramachandran left-handed helix region) that other amino acids cannot adopt. In a folded protein this is sometimes structurally significant. In an IDR, where the protein is already conformationally heterogeneous, the impact is more subtle — the ensemble's accessible conformational space shifts slightly, but no single 'wild-type' geometry is being broken.

DynaMut2's |ΔΔG| of 0.35 kcal/mol is not interpretable as a quantitative claim in this region. AlphaMissense's 0.100 score (well below the 0.564 pathogenic threshold) considers the variant likely benign.

The conflicting ClinVar classifications — some pathogenic, some uncertain — likely reflect that this variant has been observed in patients with Wolfram-spectrum disease but causal contribution has not been firmly established. The mechanism, if pathogenic, would likely involve IDR-mediated phase separation, partner binding through disordered regions, or context-dependent functional disruption — none of which AlphaMissense's training reliably captures.