A framework to identify physiological responses in microarray-based gene expression studies: selection and...

Area of Science:

• Genomics
• Bioinformatics
• Systems Biology

Background:

• Whole genome microarray studies often miss small, biologically significant gene expression changes.
• Existing pathway analysis tools are limited by their focus on annotated genes, ignoring substantial data.

Purpose of the Study:

• To develop a structured framework for identifying informative genes and interpreting biological relevance in microarray data.
• To address limitations of current methods by incorporating all genes and gene-gene interactions.

Main Methods:

• A framework involving gene ranking, selection, grouping, and biological interpretation was developed.
• Random Forests (RF) were used for gene ranking and selection, considering gene-gene interactions.
• A permutation approach established an objective threshold for gene selection.

Main Results:

• The framework successfully identifies both major and minor gene expression changes across human, mouse, and rat whole genome arrays.
• Random Forests combined with self-organizing maps detected coordinated, small gene expression responses in unannotated genes.
• The approach accounts for all genes, including those without full functional annotation.

Conclusions:

• The presented framework offers a flexible and comprehensive method for the biological interpretation of microarray data.
• It enhances the ability to detect subtle yet important biological signals by considering all genes and their interactions.
• This structured approach provides an objective method for gene selection and interpretation in complex genomic studies.