Microarray gene expression analysis
RNA quality was determined by analysis on the Agilent 2100 Bio-analyzer, and all samples had a RNA integrity number (RIN) >8. Total RNA (100ng) was labelled, and processed automatically on an HT MG-430 PM array plate using the Affymetrix GeneTitan system in the St. Jude microarray core according to the manufacturer's instructions. Detailed protocols are available on-line from Affymetrix, or upon request from the authors.
The more than 500.000 probes on the HT MG-430 PM array were redefined according to Dai et al. [1] utilizing current genome information. In this study probes were reorganized based on the Entrez Gene database, build 37, version 1 (remapped CDF v13).
http://brainarray.mbni.med.umich.edu/Brainarray/Database/CustomCDF/13.0.0/entrezg.asp
Normalized expression estimates were generated from the raw intensity values using the RMA algorithm in the Bioconductor library AffyPLM using default settings [2]. Differentially expressed probesets were identified using linear models, applying moderated t-statistics that implement empirical Bayes regularisation of standard errors [3]. To adjust for both the degree of independence of variances relative to the degree of identity and the relationship between variance and signal intensity, the moderated t-statistic was extended by a Bayesian hierarchical model to define a intensity-based moderated T-statistic (IBMT) [4]. P-values were corrected for multiple testing using a false discovery rate method [5]. Probesets that satisfied the criterion of FDR < 5% (q-value < 0.05) were considered to be significantly regulated. Changes in gene expression were related to functional changes using gene set enrichment analysis (GSEA) [6]. This analysis assesses the over- and under-representation of a known set of genes, e.g. a metabolic pathway or signal transduction route, within the list of probed genes. In this study gene sets were derived from the Gene Ontology, KEGG, NCI, PFAM and Biocarta pathway databases. Enrichment Map, a network-based visualization methods, was used for interpretation of the GSEA results [7]. Only gene sets consisting of more than 10 and less than 500 genes were taken into account in the GSEA. The enrichment map was generated only with the gene sets that passed the significance threshold of p-value < 0.005 and similarity cut-off value of 0.6, resulting in a network of 226 nodes (gene sets) and 670 edges (interactions). All singletons were removed to create the final gene set interaction network.
The raw data files are available through the Gene Expression Omnibus (GEO), accession number GSE25205.
Please contact Rinke Stienstra (rinke.stienstra
Literature