CASAVA

CASAVA predicts disease category–specific risk for non-coding sequence variants across the human genome, producing locus-level scores across 24 major disease categories.

Key Features:

• Ensemble learning framework: Implements an ensemble learning framework to integrate input features for prediction.
• GWAS training data: Uses disease-associated variants identified by Genome-Wide Association Studies (GWAS) as training data.
• Sequencing-based features: Utilizes a diverse set of sequencing-based genomics and epigenomics profiles as input features.
• Locus-level scoring: Generates scores at specific genomic loci for non-coding variants quantifying disease risk.
• Disease category coverage: Produces risk predictions across 24 major categories of diseases.
• Disease-specific outputs: Provides both disease-specific and disease category–specific risk scores.

Scientific Applications:

• Variant–disease association discovery: Reveals variant–disease associations, exemplified by analyses involving MHC2TA and immune system diseases.
• Non-coding variant prioritization: Prioritizes non-coding sequence variants genome-wide for studies of functional impact.
• Complex disease genetics: Supports investigation of the genetic underpinnings of complex diseases through disease-specific risk scoring.

Methodology:

Trains an ensemble model using GWAS-identified disease-associated variants as training labels and sequencing-based genomics and epigenomics profiles as input features, then computes locus-level scores representing disease-specific and disease category–specific risk.