GMDR

GMDR identifies gene-by-gene and gene–environment interactions in genetic studies using a generalized, nonparametric, model-free multifactor dimensionality reduction approach to detect epistasis while accommodating covariates and diverse phenotype types.

Key Features:

• Model-free nonparametric approach: Provides a nonparametric, model-free alternative to linear or logistic regression to mitigate the curse of dimensionality in high-dimensional interaction analyses.
• Adjustment for covariates: Allows inclusion and adjustment of both discrete and quantitative covariates in interaction analyses.
• Applicability to diverse phenotypes and sample structures: Handles dichotomous and continuous traits and supports unrelated case-control studies, family-based samples, pooled unrelated and family samples, and multivariate phenotypes.
• Enhanced detection capability: Demonstrated superior performance in detecting epistatic loci via computer simulations and has identified joint gene actions such as between CHRNB4 and NTRK2 in nicotine dependence.
• Permutation testing and empirical p-values: Implements permutation testing strategies to evaluate empirical p-values for robust statistical assessment.
• Data management, preprocessing, and scalability: Includes data management and preprocessing features and scales to large-scale or genome-wide analyses according to available computational resources.

Scientific Applications:

• Complex trait genetics: Detection and characterization of nonlinear gene-gene and gene-environment interactions underlying complex traits.
• Addiction genetics: Analysis of genetic interactions contributing to nicotine dependence and related phenotypes.
• Disease susceptibility studies: Identification of epistatic loci and interaction effects influencing disease risk.
• Evolutionary and multivariate trait analysis: Investigation of multifactor interactions in evolutionary biology and multivariate phenotype contexts.

Methodology:

Generalized multifactor dimensionality reduction framework that is nonparametric and model-free; inclusion and adjustment of discrete and quantitative covariates; support for dichotomous and continuous phenotypes and unrelated, family-based, pooled, and multivariate samples; data management and preprocessing for large-scale/genome-wide analyses; permutation testing to obtain empirical p-values; performance evaluated by computer simulations.