GWASinlps

GWASinlps performs iterative SNP selection using non-local priors to identify associated variants and quantify uncertainty in genome-wide association studies.

Key Features:

• Non-Local Priors: Applies non-local priors to enhance parsimonious uncertainty quantification during SNP selection.
• Iterative Variable Selection Framework: Implements an iterative screen-and-select framework that evaluates response-predictor and response-response associations within hierarchical structures.
• Structured Screen-and-Select Strategy: Employs hierarchical screening to prioritize variables based on direct and indirect associations to reduce false discoveries.
• Computational Efficiency: Integrates association learning techniques within the iterative framework to balance computational efficiency with robust statistical inference for large-scale genomic data.
• Empirical Power Analysis: Provides empirical power analysis to inform study design and assess detection power.

Scientific Applications:

• GWAS SNP identification: Identification of single nucleotide polymorphisms associated with complex traits in genome-wide association studies.
• Phenotype analysis (human height): Application to diverse phenotypes, exemplified by analysis of human height data.
• Linkage disequilibrium handling: Handling realistic linkage disequilibrium structures in genotype data to improve inference.
• Effect size estimation and error minimization: Improving effect size estimation accuracy while minimizing false discoveries and estimation error.

Methodology:

Uses non-local priors within an iterative screen-and-select algorithm with hierarchical screening that evaluates response–predictor and response–response associations, integrates association learning techniques, and performs empirical power analysis and extensive simulation studies comparing performance to frequentist and Bayesian variable selection methods using metrics including true positive rate, false discovery rate, mean squared error, and effect size estimation accuracy.