BiologybioRxiv
Ranked by formulaVerdict: NotableAlphaGenome informed machine learning for pathogenicity prediction of germline mutations
Accurate interpretation of germline variants remains a central challenge in clinical genomics.
Key numbers
- 1 x 10-300
- 95% CI 0.9731-0.9775
- 95% CI 0.1489-0.1589
Caveats
- Preprint; not yet peer reviewed.
VerdictWorth a reader's time today.
Abstract
Accurate interpretation of germline variants remains a central challenge in clinical genomics. Existing in silico pathogenicity predictors often provide incomplete coverage or restricted to missense variants from the coding region, leaving the variants from splice sites, -UTR, and intronic regions under-scored. We present AGVarPred, a gradient-boosted classifier designed to predict the pathogenicity of germline mutants, with a broad consequence-class coverage across SNVs and indels. Here, the population allele frequency from gnomAD, functional consequence annotations from Ensembl VEP, and cell-type-specific functional genomic scores from the AlphaGenome platform were integrated to develop the classifier. Using a gene-level split of 134,002 high-confidence ClinVar variants across 9,222 genes, AGVarPred achieved an AUC-ROC of 0.9753 (95% CI 0.9731-0.9775) and a AUC-PR of 0.9606 on 12,385 scored internal test variants. A paired DeLong analysis showed that the full model significantly outperformed an allele-frequency-plus-VEP model ({Delta}AUC = 0.1539, 95% CI 0.1489-0.1589; p < 1 x 10-300), supporting a substantial incremental contribution from AlphaGenome-derived features. Performance remained strong when restricted to SNVs (AUC-ROC 0.9558) and was also high for indels (AUC-ROC 0.9532). On five external benchmarks, AUC-ROCs ranged from 0.528 on a functional-assay-derived benchmark to 0.977 on a clinically curated deafness database, reflecting its dependence on benchmark composition and label source. A separate analysis of the VIP Database, whose pathogenic labels are ClinVar-derived, showed consistent performance (AUC-ROC of 0.892) in held-out genes. AGVarPred scored 100% of evaluated variants in the external benchmarks, whereas AlphaMissense and REVEL covered only small fractions of non-missense variants. Model interpretation indicated that gnomAD allele frequency was the dominant individual predictor; the 111 AlphaGenome features provided complementary functional signal, with splicing and chromatin assays contributing to the largest share of the AlphaGenome SHAP mass. A web-based interface for AGVarPred is available at https://pankaj357.github.io/AGVarPred/ to facilitate the submission of normalized GRCh38 germline variants and retrieval of pathogenicity predictions using the frozen AGVarPred model.
The editor's rubric
- Leverage4 / 5
- A general-purpose tool used across several fields (Adam, ResNet, LoRA, next-generation sequencing).
- Magnitude3 / 5
- Large gain: roughly 2x, or a clear new state of the art on a hard, unsaturated problem.
- Evidence3 / 5
- Solid: multiple benchmarks or cohorts, ablations, fair baselines, released code or data.
- Novelty2 / 5
- A new combination of known ideas.
- Trajectory3 / 5
- A clear path to scale.
- Stakes2 / 5
- Benefits a professional community (practitioners, clinicians, engineers).
Editor’s rationale
Heuristic triage from title and abstract text only, not a reading of the paper. Cues found: method (we propose); breadth (many tasks, programmable); gains (outperforms); verification (confidence interval, multiple benchmarks, held-out test); stakes (genetic disease). Red flags: weak evidence (in silico only).
How the score was computed
Desks place stronger verdicts first, then use this score to order work within the same verdict.
- Merit
- 5.9 / 10
- Adjusted merit
- 4.9 / 10
- Attention
- 0%
- Freshness
- 88%