Computational protocol: Alternative transcription of a shorter, non-anti-angiogenic thrombospondin-2 variant in cancer-associated blood vessels

Similar protocols

Protocol publication

[…] In the human exon 1.0 ST array (Affymetrix), each exon is represented by a probeset comprising at least four oligonucleotides. Probesets with the highest level of supportive evidence (core probesets (n = 284,000)) that target all exons of transcripts annotated by Refseq were used for exon analysis. Raw fluorescence intensity value summarization is described extensively in Supplementary Material and Methods under: Human Exon 1.0 ST probe summarization.Alternative splicing was investigated using three parallel analyses. To detect differential splicing/transcription, the effect of differential gene expression needs to be removed. A commonly used strategy to achieve this is to calculate a splicing index (SI, []) for each exon, which is denoted as the log 2 value of the ratio of expression levels of individual exons to overall gene level expression. The SI was used as a measure for differential splicing between two groups. An SI value of 0 then indicates equal expression of a particular exon in a transcript. SI values of 1 and −1 were used as arbitrary thresholds. We used this SI in an analysis of variance (ANOVA) model named MIDAS [], using Genespring Gx 12 (Agilent) and Altanalyze [] to test the null hypothesis that no alternative splicing occurs for a particular exon. For the sake of clarity, SI when used in the MIDAS ANOVA model is called MIDAS score. Exons with a MIDAS p value <0.05 were considered significant for exon inclusion/exclusion. Second, ANOSVA [] as performed on raw exon values using the Partek Genomics Suite (Partek Incorporated). This method directly uses the probeset intensity signals of individual exons to detect differences in the expression of exons using an ANOVA model. Exons with an ANOSVA value of p < 0.01 were considered significant for exon inclusion/exclusion. A third analysis was performed using the FIRMA [] using the statistical language R/Bioconductor []. FIRMA predicts alternative splicing of an exon based on the distance from the transcript expression estimate produced by the RMA model. The FIRMA score is a measure of this distance. A FIRMA score of 1 indicates that no splicing has occurred. FIRMA scores were compared using Bayesian statistics. FIRMA was performed using the default implementation of the aroma.affymetrix package in R together with logarithmic transformation, as described in the human exon array analysis vignette [, ]. A custom chip description file covering the exon array core probesets was used (HuEx-1_0-st-v2,coreR3,A20071112,EP.cdf, created by E Purdom). After obtaining the FIRMA scores per probeset per sample, the mean FIRMA score in both experimental groups and the difference between the groups was calculated. Finally, the probesets were mapped to transcript clusters and the transcripts were sorted by the FIRMA score. The FIRMA scores were also subjected to statistical testing using Bayesian statistics as implemented in the limma [] statistical package to identify consistent splicing differences across samples. Exons whose FIRMA scores were lower than p < 0.001 were considered significant for exon inclusion/exclusion. A detailed description of the analysis is provided in Supplementary Material and Methods under: Comparison of different exon array methods of analysis. […]

Pipeline specifications

Software tools GeneSpring GX, AltAnalyze, ANOSVA, Partek Genomics Suite, FIRMA, Aroma.affymetrix, limma
Application Gene expression microarray analysis
Diseases Urinary Bladder Neoplasms, Carcinoma, Squamous Cell, Neoplasms, Corneal Neovascularization