Nested, non-parametric, correlative analysis of microarrays for heterogenous phenotype characterization

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-dimensional genomic data presents challenges in identifying relevant genes for phenotype characterization.
• Existing methods often rely on individual gene expression differences, potentially overlooking complex biological patterns.

Purpose of the Study:

• To develop a non-parametric approach for selecting candidate genes based on gene signal profiles.
• To reliably characterize phenotypes using high-dimensional data from limited samples.
• To differentiate biological variations from artifacts in gene expression data.

Main Methods:

• A novel closeness measure based on gene signal profiles (functionals) rather than isolated differences.
• Sequential examination of gene set significance: all arrayed genes, candidate genes, and individual genes.
• Utilizing U-statistics within a sample pair analysis framework.
• Application to a microarray experiment for skin cancer subtype identification.

Main Results:

• The proposed measure effectively separates biological variation from artifactual variation (e.g., tissue effects, signal calibration).
• The method successfully identifies sets of genes that characterize specific phenotypes, such as skin cancer subtypes.
• Demonstrated ability to assess the direction of gene expression (over- or under-expressed).

Conclusions:

• The non-parametric approach provides a robust method for qualitative gene selection in high-dimensional genomic studies.
• Gene signal profile analysis offers a powerful alternative to distance-based methods for phenotype characterization.
• This method enhances the reliability of identifying biologically relevant genes for disease subtyping.