OrdinalGWAS.jl

OrdinalGWAS.jl performs genome-wide association analysis of ordered categorical phenotypes using an ordered multinomial regression to increase statistical power and handle biobank-scale datasets.

Key Features:

• Ordered Multinomial Regression: Employs an ordered multinomial regression model for phenotypes with natural orderings (e.g., undiagnosed, pre-disease, mild, moderate, severe).
• Increased Analytical Power: Improves statistical power relative to dichotomization or treating ordinal data as quantitative, yielding more robust association signals.
• Efficient Computation: Implements efficient algorithms for computing test statistics to scale analyses to large biobank datasets.
• Biobank-Scale Application: Demonstrated applicability to large datasets such as the UK Biobank, including analysis of hypertension as an ordinal trait.
• Validation in External Study: Confirmed previously identified genetic associations with greater significance than binary case-control analyses in the COPDGene study.

Scientific Applications:

• Clinical Subtypes: Analysis of phenotypes categorized from comprehensive clinical information that take ordered forms.
• Derived Phenotypes from EHR: Genome-wide analysis of traits generated by phenotyping algorithms applied to electronic health records (EHR).
• Biobank-Scale GWAS: Application to large-scale genomic studies such as UK Biobank and COPDGene for ordinal traits, including hypertension severity.

Methodology:

Implements an ordered multinomial regression model and efficient algorithms for computing test statistics for GWAS on ordinal phenotypes.