Computational evidence supporting VUS to Likely Pathogenic reclassification for NM_153700.2:c.4976A>C p.(Glu1659Ala)
Every possible amino acid substitution at position 1659 is predicted Likely Pathogenic. This position is structurally invariant: any change breaks the protein.
E1659 is 100% conserved across all tested mammals, spanning ~80 million years of evolution. The surrounding motif PEIFTEIGTIAAG is identical in every species.
| Species | Position | Residue | Context |
|---|---|---|---|
| Human | 1659 | E | PEIFTEIGTIAAG |
| Mouse | 1693 | E | PEIFTEIGTIAAG |
| Rat | 1693 | E | PEIFTEIGTIAAG |
| Cow | 1647 | E | PEIFTEIGTIAAG |
| Green monkey | 1659 | E | PEIFTEIGTIAAG |
| Pig | 1650 | E | PEIFTEIGTIAAG |
| Dog | 1649 | E | PEIFTEIGTIAAG |
| Bat | 1646 | E | PEIFTEIGTIAAG |
| Bear | 1643 | E | PEIFTEIGTIAAG |
9/9 species conserve Glutamic acid (E) at this position. The surrounding 13-residue motif is identical across all tested mammals. This level of conservation strongly suggests functional importance. Note: this conservation data is integrated within the REVEL score (via GERP++, SiPhy, phyloP) rather than as a separate PP1 criterion, as PP1 is reserved for co-segregation in affected family members. Data source: UniProt ortholog sequences, motif-based alignment.
Stereocilin (Q7RTU9, 1775 aa) from AlphaFold v6. Position E1659 highlighted in magenta. Drag to rotate, scroll to zoom.
Color: pLDDT confidence (blue=high, red=low)
Glutamic acid side chain shown as sticks
STRC has a nearly identical pseudogene (STRCP1) located adjacent on chromosome 15q15.3. This causes most standard computational tools to fail or return unreliable results for STRC variants:
AlphaMissense is uniquely valuable for STRC because it predicts pathogenicity from protein structure, bypassing the sequence-alignment step where pseudogene STRCP1 causes other tools to fail. REVEL (0.789) and CADD (PHRED 27.5) also provide concordant predictions, using ensemble approaches that partially mitigate this issue.
13 of 15 predictors classify E1659A as damaging. Scores aggregated from 22 tools tested on this variant.
Overwhelming computational consensus across meta-predictors, single-tool predictors, and protein stability analysis. The only partial outlier is PrimateAI (0.559, Tolerated), explained by the STRC pseudogene confounding primate alignments.
| Tool | Score | Verdict |
|---|---|---|
| AlphaMissense | 0.9016 | Likely Pathogenic |
| ClinPred | 0.9869 | Pathogenic |
| DANN | 0.9946 | Highly deleterious |
| FATHMM-MKL | 0.9748 | Damaging |
| MetaRNN | 0.8552 | Damaging |
| REVEL | 0.789 | Pathogenic range |
| VEST4 | 0.590 | Moderate signal |
| BayesDel | 0.2255 | Damaging |
| CADD | PHRED 25.5 | Top 0.3% |
| SIFT | 0.0 | Deleterious |
| PolyPhen-2 | 0.991 | Probably damaging |
No clinical training data used. Models predict variant effect purely from evolutionary protein sequences.
| Model | Log-likelihood delta | Verdict |
|---|---|---|
| ESM-1v | delta -2.733 | Damaging (zero-shot) |
| ESM-2 (650M) | delta -1.718 | Damaging (zero-shot) |
Glutamic acid (E) is the most preferred residue at position 1659. Alanine (A) ranks 3rd worst of 20 possible substitutions. Ran locally on Apple M5 Max via MPS.
| Tool | Score | Verdict |
|---|---|---|
| DynaMut2 | -0.913 kcal/mol | Destabilizing |
| AlphaFold DB | pLDDT 95.69 | High confidence at E1659 |
| Tool | Score | Verdict |
|---|---|---|
| SpliceAI | 0.00 | Not splice-disrupting |
| AlphaGenome | Quantile 0.997+ | Splice context significant |
| Database | Result | Significance |
|---|---|---|
| gnomAD | Absent | 0 / 251,000+ (PM2) |
| BRAVO / TOPMed | Absent | Not in deep WGS |
| dbSNP | No rsID | Novel variant |
| ClinVar | Not submitted | Pending submission |
| LOVD | Absent | Not in hearing loss DB |
All tool pages with methodology and source links available at /tools. Scores extracted via Ensembl VEP, dbNSFP v4 (myvariant.info), and direct tool APIs.
| Criterion | Strength | Evidence |
|---|---|---|
| PM3 | Moderate | Detected in trans with pathogenic whole-gene deletion (confirmed paternal) |
| PP3_Moderate | Moderate | AlphaMissense 0.9016 + REVEL 0.789 + CADD PHRED 27.5 concordant (Pejaver 2022 threshold) |
| PM2_Supporting | Supporting | Absent from gnomAD (0 alleles in 251,000+ individuals) |
Note: conservation data (E1659 100% conserved, 9 mammals) is already captured within PP3_Moderate (REVEL integrates conservation via GERP++/SiPhy/phyloP). PP1_Supporting was removed from this analysis — PP1 is reserved for co-segregation in affected family members, which cannot be applied here (no other affected relatives with hearing loss).
2 Moderate + 1 Supporting = VUS (Variant of Uncertain Significance). Per the Holt Lab framework (Boston Children's Hospital / Harvard Medical School), this subclassifies as VUS-high — a category recognized by many clinical trial sites for recessive disease enrollment when one allele is confirmed pathogenic.