High-dimensional covariance matrices tests for analyzing multi-tumor gene expression data.

Area of Science:

• Genomics
• Statistical Genetics
• Bioinformatics

Background:

• Gene set analysis in microarray data often requires characterizing intra-subject variation.
• Understanding intra-subject (e.g., tumor type) variation is crucial for accurate gene expression profiling.
• Existing methods may not fully capture the complexities of multi-set data within subjects.

Purpose of the Study:

• To develop and validate statistical tests for assessing intra-subject variation in gene expression data.
• To test the assumption of intra-subject variation across different sets (e.g., tumor types) within subjects.
• To evaluate the properties of these tests in both theoretical and empirical settings.

Main Methods:

• Development of multi-set sphericity tests.
• Development of multi-set identity of covariance structure tests.
• Application and validation using The Cancer Genome Atlas (TCGA) data.

Main Results:

• The proposed tests for multi-set sphericity and covariance structure identity demonstrate good statistical properties.
• Theoretical and empirical studies confirm the reliability of the developed tests.
• Analysis of TCGA data provided insights into covariance structures of gene expression across tumor types.

Conclusions:

• The novel statistical tests effectively characterize intra-subject variation in gene expression data.
• These methods enhance the analysis of multi-set microarray data, particularly in cancer genomics.
• The findings contribute to a better understanding of gene expression patterns within and across different tumor types.