Analysis of high dimensional data using pre-defined set and subset information, with applications to genomic data

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• Genomic data is often structured into sets and subsets (e.g., biological pathways).
• Existing methods for differential gene expression analysis often ignore subset information and gene correlations.
• Ignoring these factors can lead to reduced statistical power and increased false positive rates.

Purpose of the Study:

• To develop a novel methodology for analyzing differential gene expression that incorporates subset information and inter-variable dependencies.
• To enable the identification of differentially expressed gene sets and their significant subsets.

Main Methods:

• A multiple testing-based approach is proposed.
• The methodology leverages information on biologically relevant subsets within predefined gene sets.
• It exploits the underlying correlation structure among variables (genes).

Main Results:

• The new methodology allows for the determination of differential expression at both the set and subset levels.
• It effectively utilizes information on subsets and variable dependencies.

Conclusions:

• The proposed method is easy to implement and robust to various data distributions and covariance structures.
• It controls the family-wise error rate (FWER) effectively.
• Applicable to various high-dimensional data types beyond microarray, including mRNA sequencing and CpG methylation data.