Franny Jongbloed

72 CHAPTER 3 consisted of 45,141 probes 33 . Both raw and normalized microarray data and their MIAME compliant metadata were deposited at the Gene Expression Omnibus (GEO) database, with number GSE72484 (www.ncbi.nlm.nih.gov/geo ). Statistical analyses For each set of parameters means and standard errors of the mean were computed. All standard statistical tests were performed using SPSS version 21 for Windows software (Statistical Package for Social Sciences, Chicago, IL) and GraphPad Prism (GraphPad Software Inc., version 5.01). A P- value <0.05 was considered to be significant. Microarray analyses were performed using the free software package R (R foundation). Gene expression profiles were compared using the Linear Models for Microarray Data (limma) method with correction for multiple testing using the false discovery rate (FDR) according to Benjamini and Hochberg 34 . Fold changes were expressed as the geometric mean per diet group against the corresponding AL fed control group, and cutoff values for a significant difference were put at FDR <5%. Functional annotation and analyses were performed using the Ingenuity software (http://www.ingenuity.com/products/ipa ). Inhibition or activation prediction of the upstream transcription regulators (upstream analysis) was predicted with Ingenuity software by calculating statistical z-scores based on the observed gene expression changes in our dataset. Via z-scores, the chance of significant prediction based on random data is reduced (http://ingenuity.force.com/ipa/articles/Feature_Description/Upstream- Regulator-Analysis). Cutoff values for a significant activation or inhibition were put at a z-score of ≥2 or ≤-2, respectively.