SeverYX: An ensemble framework for robust variant severity classification

SeverYX classifies MANE-transcript variants into High/Medium/Low with 98.7% concordance to ClinVar after adding AlphaMissense and REVEL, improving MCC to 0.908.

• Why it matters: Clinical variant interpretation must reconcile conservation, splicing, and functional-impact predictors that disagree or leave gaps, especially across different molecular consequences and evidence quality.
• What they did: The rule-based, effect-aware ensemble was benchmarked on 270,208 MANE-transcript variants from ClinVar, using molecular consequence, review status, and haploinsufficiency, with threshold optimization and added missense predictors.
• The result: Compared with the baseline (97.6% concordance), the optimized model reduced benign High calls to 0.4% while High sensitivity for pathogenic variants dropped to 88.3%, and MCC increased from 0.895 to 0.908.

AI summary of the abstract · may contain errors · not medical advice · read the paper.