Correlation and variation-based method for identifying reference genes from large datasets.

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Reference genes are crucial for normalizing gene expression data.
• Existing algorithms (NormFinder, geNorm, BestKeeper) are unsuitable for large microarray datasets.
• Need for robust methods to identify reliable reference genes in high-throughput studies.

Purpose of the Study:

• Evaluate existing reference gene identification algorithms on microarray data.
• Develop novel, accurate, and efficient methods for large datasets.
• Provide a validated tool for reference gene selection.

Main Methods:

• Correlated outputs of 7 published reference gene identification methods using microarray subsets.
• Developed and evaluated 7 novel combinations of existing methods.
• Assessed computational efficiency and correlation with established indices.

Main Results:

• High correlation observed between NormFinder and geNorm indices (R(2) = 0.987).
• Lower correlation found between NormFinder and BestKeeper (R(2) = 0.489).
• Two novel methods demonstrated high correlation with NormFinder (R(2) = 0.796) and linear computational time.

Conclusions:

• Developed novel methods are effective alternatives to existing algorithms for large datasets.
• The developed methods are implemented in the OLIgonucleotide Variable Expression Ranker (OLIVER) tool.
• OLIVER provides a downloadable solution for accurate reference gene identification.