Question: Reading methylationand gene expression with only IDAT files downloaded from GEO

Question

[emiliomastriani]

Dear all, I am going to analyze some GSM from the GSE68379 from GEO. I need the gene expression from the cell lines that I will take in consideration (or the list of genes that are differentialy expressed among these cell lines). I have experience with the gene expression data for microarray analysis, but it's the first time that I work with methylated dataset. In details, I have these 2 troubles: 1. the current dataset doesn't present any differentiation by groups (normal, disease, etc) and all the cells I will consider are of the same kind (epithelial ovarian cancer), here is the question: how to make the model to produce the differentiated methylated regions? 2. supposed that I will solve the problem n.1 and get some results from the previous step obtaining a list of differentiated CpG, how to get the expression level of genes? It looks that this dataset doesn't publish this information ... or maybe that I don't know where to look.

Please guys, try your best to support me...I already spent many days without any acceptable results

Answer 1

[Kevin Blighe]

That study's data appears to be arrange differently from others. Maybe the authors did not upload it as required. For example, usually, one can download normalised data with metadata by just following this: A: Normalization of the Illumina HumanMethylation450 BeadChip data, platform GPL135

For your study of interest, it seems that you can download all necessary files manually:

• ftp://ftp.ncbi.nlm.nih.gov/geo/series/GSE68nnn/GSE68379/matrix/ (metadata)
• ftp://ftp.ncbi.nlm.nih.gov/geo/series/GSE68nnn/GSE68379/suppl/GSE68379_Matrix.processed.txt.gz (~4GB; normalised data - this actually contains 2 columns per file, it seems: 1, methylation β value; 2, detection p-value)

Once you actually obtain the data, you can then compare probes across groups via an appropriate statistical test (Student's t-test / Mann-Whitney test, Paired t-test / Wilcoxon Signed Rank test, ANOVA / Kruskal-Wallis test, etc.). To add an extra measure, obtain the difference in mean between your groups and use that as an extra filter (e.g. DiffMean = absolute(mean(group1) - mean(group2)))

Kevin