Benchmarking AlphaMissense against ClinVar for Diagnostic Interpretation of Missense Variants in Inherited Retinal Diseases
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Purpose AlphaMissense is a newer deep learning-based variant predictor that evaluates the structural consequences of missense variants, the most common pathogenic variant type in inherited retinal diseases (IRDs). This study evaluates the diagnostic utility of AlphaMissense in IRDs by assessing its concordance with ClinVar annotations and exploring how other variant-level metrics may refine its predictions. Design Cross-sectional benchmarking study using public variant resources, with a single illustrative clinical case Participants Missense variants from 107 IRD genes; one patient case undergoing long-read sequencing. Methods Pathogenicity scores from AlphaMissense were extracted from 128,248 variants present in both IRD genes and gnomAD. Among these, 4,204 had definitive ClinVar classifications and were used to calculate AlphaMissense specificity, sensitivity, and false discovery rate (FDR). Population-based metrics including allele frequency, homozygote count, and CADD score were analyzed to identify salient features that would be associated with discordance. Long-read sequencing was carried out in a monoallelic ABCA4 patient with late-onset macular dystrophy for phased variant analysis Main Outcome Measures Concordance between AlphaMissense predictions and ClinVar annotations were used to calculate sensitivity, specificity, and FDR. Variant level metrics between discordant variants. Case-based reclassification of hypomorphic variants with long-read sequencing. Results AlphaMissense achieved a specificity of 94.1% and sensitivity of 79.4% in IRD genes, with specificity reaching 100% in A BCA4, USH2A, RPGR, and PRPH2 , which are four of the most common IRD genes. The FDR was 9.6%. AlphaMissense underperformed in predicting hypomorphic variants, particularly in ABCA4 -associated Stargardt disease. Variant-level metrics were effective in identifying false negatives. In a clinical case, phased variant analysis identified a potential hypomorphic ABCA4 variant predicted as benign by AlphaMissense Conclusions AlphaMissense demonstrates high specificity for pathogenicity prediction in IRD-associated genes, however, its reduced sensitivity, as seen in hypomorphic variants, suggests a need to incorporate population and functional metrics scores may improve classification accuracy, especially long-read sequencing enables phased variant analysis.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it