Related Experiment Video
Updated: Aug 10, 2026
13:14
Global Gene Expression Analysis Using a Zebrafish Oligonucleotide Microarray Platform
Published on: August 10, 2009
How to decide? Different methods of calculating gene expression from short oligonucleotide array data will give
Frank F Millenaar1, John Okyere, Sean T May
1Plant Ecophysiology, Institute of Environmental Biology, Faculty of Science, Utrecht University, Sorbonnelaan 16, 3584 CA Utrecht, The Netherlands. f.f.millenaar@bio.uu.nl
BMC Bioinformatics
|March 17, 2006
Summary
Comparing gene expression algorithms reveals significant variability. RMA offered the most reproducible results, but not all its findings were validated by Real Time RT-PCR.
Area of Science:
- Bioinformatics
- Computational Biology
- Molecular Biology
Background:
- Short oligonucleotide arrays are established tools for transcript profiling.
- Analysis typically relies on Affymetrix's Microarray Analysis Suite (MAS) or GeneChip Operating Software (GCOS).
- Emerging analysis methods necessitate evaluation for reliability.
Purpose of the Study:
- To evaluate and compare the reliability of different gene expression analysis algorithms.
- To assess algorithm performance using a common Arabidopsis data set.
- To validate findings with Real Time RT-PCR.
Main Methods:
- Six algorithms were compared: MAS5, dChip PMMM, dChip PM, RMA, GC-RMA, and PDNN.
- Analysis was performed on a single Arabidopsis data set.
- Real Time RT-PCR was used for validation.
Main Results:
- Significant differences in calculated gene expression levels were observed across algorithms.
- Only 27-36% overlap in identified differentially regulated genes between methods.
- 47.5% of genes/probe sets showed good correlation between mismatch and perfect match intensities.
Conclusions:
- RMA demonstrated the highest reproducibility and correlation with Real Time RT-PCR for commonly identified genes.
- Real Time RT-PCR validation failed for many genes solely identified by RMA.
- Mismatch subtraction may underestimate expression values; no algorithm detected genes below 1 pmol.
