Related Experiment Videos
Profound effect of normalization on detection of differentially expressed genes in oligonucleotide microarray data
Reinhard Hoffmann1, Thomas Seidl, Martin Dugas
1Department of Bacteriology, Max von Pettenkofer Institut, Pettenkoferstrasse 9a, 80336 Munich, Germany. r_hoffmann@m3401.mpk.med.uni-muenchen.de
Genome Biology
|August 20, 2002
Summary
Normalization methods significantly impact the detection of differentially expressed genes from oligonucleotide microarrays. Consistent normalization is crucial for reproducible results, more so than the statistical analysis method used.
Area of Science:
- Genomics
- Bioinformatics
Background:
- Oligonucleotide microarrays are used to measure mRNA transcript abundance.
- Numerous normalization and differential gene expression detection methods exist.
- The comparative impact of these methods on gene detection is not well understood.
Purpose of the Study:
- To evaluate the impact of different normalization and statistical analysis methods on detecting differentially expressed genes.
- To assess the concordance between various analysis pipelines.
Main Methods:
- Four normalization methods were combined with three statistical algorithms.
- Analysis was performed on a prototype dataset.
- Gene lists and probability rankings were compared.
Main Results:
- The number of differentially expressed genes identified varied by approximately threefold across methods.
- High concordance in results was achieved only when using identical normalization procedures.
- Normalization method had a greater impact than the statistical algorithm.
Conclusions:
- Normalization profoundly influences the detection of differentially expressed genes.
- The choice of normalization method is critical for reproducible gene expression analysis.
- There is a need for algorithms that utilize more array-derived information beyond gene expression values.