Statistical approach for selection of biologically informative genes.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Statistical Genetics

Background:

• High-dimensional gene expression data analysis requires effective gene selection techniques.
• Existing methods often prioritize statistical relevance over biological significance.
• A need exists for gene selection methods that are both statistically sound and biologically interpretable.

Purpose of the Study:

• To introduce Boot-MRMR, a novel statistical approach for informative gene selection.
• To evaluate Boot-MRMR against existing gene selection techniques using both statistical and biological criteria.
• To provide a practical guide for selecting informative genes in genomics studies.

Main Methods:

• Developed Boot-MRMR, a method based on maximum relevance and minimum redundancy.
• Introduced biological evaluation criteria: Gene Set Enrichment with QTL (GSEQ) and Gene Ontology (GO) similarity.
• Systematically evaluated Boot-MRMR against 12 other methods using five gene expression datasets within a multiple criteria decision-making framework.

Main Results:

• Boot-MRMR consistently selects genes with higher biological relevance compared to existing methods.
• The proposed technique demonstrates competitive performance in subject classification and computational efficiency.
• Under a multiple criteria decision-making framework, Boot-MRMR is identified as the superior method for informative gene selection.

Conclusions:

• Boot-MRMR offers a statistically sound and biologically relevant approach to informative gene selection.
• The method provides a robust alternative for analyzing high-dimensional gene expression data.
• An R package, BootMRMR, is available to facilitate the application of this technique in breeding and systems biology.