GPA-Tree

GPA-Tree prioritizes GWAS signals by integrating association mapping with a decision tree and hierarchical modeling of functional annotations to identify risk-associated SNPs and annotation combinations underlying complex traits.

Key Features:

• Decision tree and hierarchical modeling: Combines a decision tree algorithm with a hierarchical modeling framework to model combinations of functional annotations and SNP risk.
• Integration of association mapping and annotations: Integrates association mapping results with functional annotation data to inform SNP prioritization.
• Simultaneous identification: Identifies risk-associated SNPs and the key combinations of functional annotations associated with those SNPs concurrently.
• GWAS prioritization: Prioritizes GWAS results based on functional-annotation-guided evidence.
• Improved detection accuracy: Demonstrates improved detection of risk-associated SNPs compared with traditional statistical approaches in simulation studies.
• Method benchmarking: Evaluated through extensive simulation studies assessing detection of causal SNPs and recovery of true annotation combinations.
• Application to SLE data: Applied to a systemic lupus erythematosus (SLE) GWAS using GenoSkyline and GenoSkylinePlus functional annotations.
• Biological insight extraction: Identified dysregulation in blood immune cell types including primary B cells, memory helper T cells, regulatory T cells, neutrophils, and CD8+ memory T cells.
• Genetic architecture elucidation: Facilitates elucidation of genetic architecture and potential mechanisms linking SNPs to complex traits.

Scientific Applications:

• GWAS signal prioritization: Prioritizes candidate SNPs from GWAS by leveraging functional annotations to rank association signals.
• Annotation-combination discovery: Detects combinations of functional annotations that are jointly associated with trait risk.
• Disease-specific analysis: Enables analysis of disease GWAS datasets, exemplified by application to SLE with GenoSkyline and GenoSkylinePlus annotations.
• Method evaluation: Provides a framework for benchmarking statistical methods via simulation studies of detection power and annotation recovery.

Methodology:

Computational steps explicitly include a decision tree algorithm combined with a hierarchical modeling framework to integrate association mapping and functional annotation analysis, with performance assessed via extensive simulation studies.