regBase

regBase aggregates base-wise prediction scores and applies composite models to predict and functionally annotate non-coding regulatory variants across the human genome.

Key Features:

• Whole-genome base-wise aggregation: Compiles base-wise aggregations of prediction scores from a multitude of existing tools across the human genome.
• Composite prediction models: Implements three distinct composite models trained to score Functional Variants, Pathogenic Variants, and Cancer Driver Mutations.
• Diverse causality assumptions: Models are built on diverse assumptions regarding causality to tailor predictions to different variant roles.
• Enhanced performance: Demonstrates superior and stable performance in predicting non-coding regulatory variants and prioritizing functional and pathogenic SNVs as validated by independent benchmarks.
• Fine-mapping capabilities: Enables fine-mapping of causal regulatory variants at locus-specific and base-wise resolutions.

Scientific Applications:

• Annotation-based variant fine-mapping: Refines identification of causal variants within non-coding regions to support interpretation of regulatory effects on gene regulation.
• Pathogenic variant discovery: Predicts pathogenic non-coding regulatory variants to aid elucidation of genetic contributors to disease.
• Cancer driver mutation identification: Identifies candidate non-coding cancer driver mutations to inform oncogenomics research and target discovery.

Methodology:

Integrates a comprehensive set of prediction scores from existing tools via whole-genome base-wise aggregation and trains three composite models (Functional, Pathogenic, Cancer Driver) using those aggregated scores, with performance assessed by independent benchmarks.