Related Experiment Videos
Effect of normalization on significance testing for oligonucleotide microarrays
Rudolph S Parrish1, Horace J Spencer
1Department of Bioinformatics and Biostatistics, School of Public Health and Information Sciences, University of Louisville, Louisville, KY 42092, USA. rudy.parrish@louisville.edu
Journal of Biopharmaceutical Statistics
|October 8, 2004
Summary
Normalization methods like quantile and RMA can inflate gene expression findings in cancer studies. This occurs due to unintended effects on experimental error variance, impacting significance tests.
Area of Science:
- Genomics
- Bioinformatics
- Biostatistics
Background:
- Normalization is crucial for reducing technical variation in gene expression data from oligonucleotide microarrays.
- Common normalization techniques include median-interquartile range (IQR) and quantile normalization.
- Gene expression values are often calculated using algorithms like MAS 5.0 and RMA.
Purpose of the Study:
- To evaluate the impact of different normalization techniques on significance testing for differential gene expression.
- To assess the effect of normalization on experimental error variance in microarray data.
- To compare the performance of MAS 5.0 and RMA algorithms, and median-IQR and quantile normalization methods.
Main Methods:
- Applied median-IQR and quantile normalization methods to a prostate cancer dataset.
- Utilized MAS 5.0 and RMA algorithms for gene expression value calculation.
- Performed paired-t significance tests and adjusted for multiple testing.
Main Results:
- Normalization methods can substantially inflate the number of significant differentially expressed genes.
- This inflation is primarily caused by an unintended effect on experimental error variance.
- The RMA method and quantile normalization showed a greater impact on error variance compared to MAS 5.0 and median-IQR, respectively.
Conclusions:
- Careful consideration of normalization methods is essential to avoid overestimation of differential gene expression.
- The choice of normalization algorithm and method can significantly influence the reliability of statistical findings.
- Further research is needed to understand and mitigate the impact of normalization on error variance in high-density oligonucleotide array analysis.