Related Experiment Video
Updated: Jun 23, 2026
05:22
Analyzing Multifactorial RNA-Seq Experiments with DiCoExpress
Published on: July 29, 2022
Ranking differentially expressed genes from Affymetrix gene expression data: methods with reproducibility,
Koji Kadota1, Yuji Nakai, Kentaro Shimizu
1Graduate School of Agricultural and Life Sciences, The University of Tokyo, 1-1-1 Yayoi, Bunkyo-ku, Tokyo 113-8657, Japan. kadota@bi.a.u-tokyo.ac.jp
Algorithms for Molecular Biology : AMB
|April 24, 2009
Summary
Selecting the right preprocessing and gene ranking methods improves microarray analysis. Fold change-based methods, specifically Rank Products (RP) or Weighted Average Difference (WAD), enhance reproducibility for identifying differentially expressed genes (DEGs).
Area of Science:
- Bioinformatics
- Genomics
- Statistical Analysis
Background:
- Microarray analysis requires selecting preprocessing algorithms and gene ranking methods to identify differentially expressed genes (DEGs).
- Previous recommendations focused on sensitivity and specificity, but reproducibility is also crucial for robust DEG identification.
Purpose of the Study:
- To evaluate combinations of preprocessing algorithms and gene ranking methods for their impact on reproducibility in microarray data analysis.
- To identify optimal methods that enhance sensitivity, specificity, and reproducibility in DEG identification.
Main Methods:
- Compared eight gene ranking methods (WAD, AD, FC, RP, modT, samT, shrinkT, ibmT) with six preprocessing algorithms (PLIER, VSN, FARMS, mmgMOS, MBEI, GCRMA).
- Evaluated 36 real experimental datasets using the area under the receiver operating characteristic curve (AUC) for sensitivity and specificity.
- Assessed reproducibility using percentages of overlapping genes (POGs) across different sites from the MicroArray Quality Control (MAQC) project datasets.
Main Results:
- Rank Products (RP) performed well with VSN, FARMS, MBEI, and GCRMA preprocessing.
- Weighted Average Difference (WAD) performed well with mmgMOS preprocessing.
- Fold change-based methods (WAD, AD, FC, RP) demonstrated higher reproducibility (higher POGs) compared to t-statistic-based methods.
- WAD showed the highest overall reproducibility among fold change-based methods, regardless of the preprocessing algorithm used.
Conclusions:
- Selecting appropriate preprocessing and gene ranking method combinations is essential for improving sensitivity, specificity, and reproducibility in microarray analyses.
- Fold change-based methods, particularly Rank Products (RP) or Weighted Average Difference (WAD), are recommended for enhanced reproducibility in DEG identification.
