Articles | Volume 52, issue 1
Arch. Anim. Breed., 52, 65–78, 2009
https://doi.org/10.5194/aab-52-65-2009
Arch. Anim. Breed., 52, 65–78, 2009
https://doi.org/10.5194/aab-52-65-2009

  10 Oct 2009

10 Oct 2009

Effects of threshold choice on the results of gene expression profiling, using microarray analysis, in a model feeding experiment with rats

A. Hartmann1, G. Nuernberg2, D. Repsilber2, P. Janczyk5, C. Walz1, S. Ponsuksili1, W.-B. Souffrant3, and M. Schwerin1,4 A. Hartmann et al.
  • 1Research Group Functional Genomics, Research Institute for the Biology of Farm Animals (FBN), Dummerstorf, Germany,
  • 2Research Unit Genetics and Biometrics, Research Institute for the Biology of Farm Animals (FBN), Dummerstorf, Germany,
  • 3Research Unit Nutritional Physiology »Oskar Kellner«, Research Institute for the Biology of Farm Animals (FBN), Dummerstorf, Germany,
  • 4Institute of Farm Animal Sciences and Technology, University of Rostock, Rostock, Germany
  • 5Institute of Veterinary Anatomy, Department of Veterinary Medicine, Free University of Berlin, Berlin, Germany

Abstract. Global gene expression studies using microarray technology are widely employed to identify biological processes which are influenced by a treatment e.g. a specific diet. Affected processes are characterized by a significant enrichment of differentially expressed genes (functional annotation analysis). However, different choices of statistical thresholds to select candidates for differential expression will alter the resulting candidates list. This study was conducted to investigate the effect of applying a False Discovery Rate (FDR) correction and different fold change thresholds in statistical analysis of microarray data on diet-affected biological processes based on a significantly increased proportion of differentially expressed genes. In a model feeding experiment with rats fed genetically modified food additives, animals received a supplement of either lyophilized inactivated recombinant VP60 baculovirus (rBV-VP60) or lyophilized inactivated wild type baculovirus (wtBV). Comparative expression profiling was done in spleen, liver and small intestine mucosa. We demonstrated the extent to which threshold choice can affect the biological processes identified as significantly regulated and thus the conclusion drawn from the microarray data. In our study, the combined application of a moderate fold change threshold (FC≥1.5) and a stringent FDR threshold (q≤0.05) exhibited high reliability of biological processes identified as differentially regulated. The application of a stringent FDR threshold of q≤0.05 seems to be an essential prerequisite to reduce considerably the number of false positives. Microarray results of selected differentially expressed molecules were validated successfully by using real-time RT-PCR.