Filtering out Probe Ids from Illumina

Question

[questionaire]

Hello All:

I have a list of Probe Ids that I would like to find differentially expressed genes for. What I have is a complete dataset from Illumina and I want to do differential expression on the subset of this data. I have already done differential expression for the complete dataset, but for the subset I want to do analysis on the differential expression on the probes from the complete dataset. Does anyone happen to know how I can either filter out the probe ids that I don't want or use the probe ids that I seek?

Thank you in advance !!!

Answer 1

[mforde84]

I haven't tested it, but this should work.

Assuming that row name is equal to probe id, and filter is equal to a list of excluded probe ids:

eset_filtered <- eset[rownames(eset) != filter, ]

Answer 2

[WouterDeCoster]

I'm not fully aware of microarray statistics, but for as much as I know (from RNA-seq) it's not recommended to do differential expression analysis on a too small subset of your data. But I don't know about the size of your subset. That said, the most appropriate approach would probably be to do differential expression analysis on the entire dataset and filter out the results. Depending on how you select the subgroup, you can filter the results before correcting for multiple testing. Without peaking at the result before correcting...

And if you really want to filter before doing differential expression analysis, you can probably do this very easily in R.

Answer 3

[questionaire]

Hello WouterDeCoster:

Thank you so much for you input. I understand that in terms of statistics it is not recommended. However my sample size for the subset is big enough. I was wondering if you happen to know how to filter out the probe ids on R before the multiple testing. Thank you!

[WouterDeCoster]

Please use ADD COMMENT to reply to earlier answers, as such this thread remains logically structured and easy to follow.

Do you know any R programming? Because the solution is probably trivial.

[questionaire]

Hello WouterDeCoster:

I do, but I'm not sure how to go about for this one. Thank you.