Multivariate Gene Selection and Testing in Studying the Exposure Effects on a Gene Set

Area of Science:

• Genomics
• Biostatistics
• Environmental Health

Background:

• Analyzing associations between gene sets and exposures is crucial in genomics.
• Identifying specific genes affected by exposures enhances biological mechanism understanding.
• Individual gene analysis is often less powerful than gene set analysis.

Purpose of the Study:

• To develop computationally simple Canonical Correlation Analysis (CCA) based variable selection methods.
• To jointly select subsets of genes within a gene set associated with exposures.
• To propose and compare model selection criteria, including the new Correlation Information Criterion (CIC).

Main Methods:

• Sparse Outcome Selection (SOS) CCA
• Step CCA
• Development of global test procedures for exposure effects on gene sets, accounting for gene selection.

Main Results:

• Proposed methods are computationally efficient and improve upon existing methods, especially for correlated genes.
• Simulation studies demonstrate the effectiveness of SOS CCA and step CCA.
• Application to the Normative Aging DNA methylation Study to investigate airborne particulate matter effects.

Conclusions:

• The proposed CCA-based methods offer a powerful and interpretable approach for gene set-exposure association studies.
• These methods facilitate the identification of biologically relevant genes and improve the power of gene set-level tests.
• The study provides valuable tools for environmental health research and understanding DNA methylation changes.