Multivariate dimensionality reduction approaches to identify gene-gene and gene-environment interactions underlying...

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Identifying multifactor interactions is crucial for understanding complex human diseases.
• Existing methods struggle with multifaceted phenotypes and longitudinal data.
• Dimensionality reduction offers a promising avenue for detecting these interactions.

Purpose of the Study:

• To develop a multivariate approach for detecting gene-environment interactions in complex diseases.
• To enhance statistical power in identifying disease determinants.
• To address challenges posed by composite syndromes and longitudinal data.

Main Methods:

• Propose a multivariate extension of generalized multifactor dimensionality reduction (GMDR).
• Utilize multivariate generalized linear, quasi-likelihood, and generalized estimating equations models.
• Employ a two-stage testing procedure: multivariate analysis followed by univariate analysis.

Main Results:

• The proposed multivariate GMDR method substantially boosts statistical power compared to univariate methods.
• Demonstrated effectiveness through simulations and analysis of the Study of Addiction: Genetics and Environment cohort.
• Successfully handles multifaceted phenotypes and longitudinal data structures.

Conclusions:

• The multivariate GMDR provides a powerful tool for dissecting complex disease etiology.
• This method advances the identification of genetic and environmental determinants of complex diseases.
• Offers a significant improvement for analyzing high-dimensional genetic and clinical data.