Hi everybody, for an old research I measured the expression levels of about 150 genes in 30 patients. Now, I'm interested to see if there is a difference, among three groups of these patients, in the expression levels of a particular gene (only one!).
Could I use DESeq2 to do that? The problem is that I can't use a statistical model directly on the raw data, then I would firstly normalize the raw data of this gene, secondly I want to do a correct statistical test.
I thought to use DEseq2 to normalize my gene using all the 150 genes, then do DESeq2 analysis and finally extract only the result for that gene (ignoring the others).
In your opinion, is it a correct procedures to do?
Thank you very much for your help, and sorry if my english is not correct, but I do not speak it very well.
Thank you again! Bye
If you're measuring other genes via RNAseq anyway, then just use everything, run DESeq2 as normal, and look at the unadjusted p-value for the single gene of interest. You're then using the other genes for shrinkage, but since you don't care whether they're differentially expressed or not you don't end up having the multiple testing issue (i.e., you can "have your cake and eat it too").
Thank you very much for your reply!
Now, what can I do if I want to create a graph with the three expressions levels (one for each group) of this gene? In your opinion, are the fitted values (using: assays(dds)[["mu"]]) corrected to plot? Or should I use the normalized values (using: counts(dds, normalized=TRUE))? I noted that these two graphs are different each other..
There's no point in plotting the fit coefficient, plot the normalized values. There's no reason that the graphs should be the same (in fact, they rarely will be).
OK perfect! I would like to graph the results of the model, how could I do? It would be correct if I represent the mean values of the groups and the standard errors of the normalized counts?
Thanks a lot again!!
"Graph the results of the model" has no meaning, it's just one number (two if you include an error). There is absolutely no point in plotting that. Put the fit coefficient in a table or directly in the text for a single gene.
Yes, you're right. Sorry, where can I find the fit coefficient?
results(dds) or something like that.
Log in to answer this question.
