CardioBoost

CardioBoost predicts the pathogenicity of rare missense variants (allele frequency ≤0.1% in gnomAD) in genes linked to inherited cardiomyopathies and arrhythmias using a disease-specific machine-learning classifier.

Key Features:

• Disease-Specific Approach: Integrates disease-specific information into the predictive model and distinguishes between gain-of-function and loss-of-function variant consequences.
• Allele Frequency Filtering: Targets rare missense variants with an allele frequency of 0.1% or less as cataloged in gnomAD.
• High Discrimination Accuracy: Achieves precision-recall AUC of 0.91 for cardiomyopathies and 0.96 for arrhythmias, outperforming existing tools by 4–24%.
• Confidence in Variant Classification: Reaches 90.2% accuracy for cardiomyopathies and 91.9% accuracy for arrhythmias when variants are classified with >90% confidence, and more than doubles the proportion of variants classified with high confidence versus existing tools.
• Clinical Relevance: Variant classifications correlate with clinical outcomes, including disease status and severity, and associate with a 21% increased risk (95% CI: 11–29%) of severe adverse outcomes by age 60 in patients with hypertrophic cardiomyopathy.

Scientific Applications:

• Clinical Genome Interpretation: Improves discrimination between benign and pathogenic rare variants for clinical genetic interpretation.
• Variant Prioritization: Prioritizes disease-associated missense variants in cardiac genes for downstream analysis.
• Patient Stratification: Enables stratification of patient outcomes and assessment of variant-associated risk and severity in inherited cardiomyopathies and arrhythmias.
• Genetic Risk Assessment: Informs genetic risk assessment potentially relevant to clinical decision making in cardiac disease contexts.

Methodology:

CardioBoost is a machine-learning classifier trained by assessing known pathogenic versus benign variants, incorporates disease-specific information, and targets rare missense variants (allele frequency ≤0.1% in gnomAD) in genes linked to cardiac conditions.