milorGWAS

milorGWAS implements mixed logistic regression to estimate genetic variant effects in genome-wide association studies (GWAS) of binary phenotypes while accounting for population structure.

Key Features:

• Mixed Logistic Regression (MLR): The core method uses mixed logistic regression, which is more appropriate than mixed linear models for binary phenotypes and accounts for population structure when estimating variant effects.
• Fast Approximate Methods: Employs fast approximate algorithms to efficiently estimate genetic variant effects for large-scale GWAS datasets.
• Novel effect-estimation methods and comparisons: Includes two novel methods for estimating variant effects that were evaluated by simulation and compared against mixed linear models and logistic regression, with MLR showing improved performance across scenarios.
• Bias Evaluation: Methods exhibit moderate bias for large effect sizes but generally provide well-evaluated variant effect estimates under typical GWAS scenarios.
• Stratified QQ-Plot: Provides a stratified QQ-plot to diagnose p-value inflation or deflation when population strata are not clearly identified within the sample.

Scientific Applications:

• Binary trait association mapping: Identification of associations between genetic variants and binary traits in GWAS.
• Analyses with complex population structure: Analysis of datasets with population structure where traditional mixed linear models or logistic regression may produce biased results.
• Method evaluation via simulation: Simulation-based assessment and comparison of effect-estimation methods for GWAS.

Methodology:

Implements mixed logistic regression, two novel effect-estimation methods, fast approximate estimation algorithms, simulation-based evaluation and comparisons against mixed linear models and logistic regression, and generates stratified QQ-plots for p-value diagnostics.