I would like to do a differential gene expression analysis on a microarray data. From the literature, I understood, call detection (Present, Absent, Marginal) has to be done to minimize false positives. I found this tutorial for the call detection and also to find the DEGs, from .cel files.

For this particular dataset used for my study, .cel file is unavailable and only .txt file/ SOFT file is available for download.

  1. Can call detection be done from these files?
  2. Or is there any other method/ package that can be used for the same?
  • $\begingroup$ Without the .cel files one doesn't have expression data, so you can't analyse the dataset (although possibly in the .txt file there is a link to each .cel file). But you seem to be using GEO. Which is the ID of the dataset you are working with? Also showing how did you get to this point would help us to find the best way to help you. $\endgroup$ – llrs Apr 11 '18 at 14:58
  • $\begingroup$ I have edited the question with more details. $\endgroup$ – Sunbiotech Apr 12 '18 at 15:34
  • 1
    $\begingroup$ You haven't actually linked to a particular dataset, but SOFT files usually contain processed expression data, so you can skip right to DGE analysis. There should be details about how the expression data has been processed in the file metadata. There's a tutorial that goes through reading these files here $\endgroup$ – heathobrien Apr 12 '18 at 15:47
  • $\begingroup$ I was about to comment that the edit wasn't enough, can you provide a link to the dataset or I misunderstood something? $\endgroup$ – llrs Apr 12 '18 at 21:03
  • $\begingroup$ The data is accessible, look at the GEO2R page to download the dataset, under the tab "R script" you'll find the code to download the dataset. It is working for me. $\endgroup$ – llrs Apr 13 '18 at 7:43

With things like MAS5, presence/absence calls were done using the P/M/A probe intensities and corresponded more to a "signal is reliable" p-value than literal presence/absence. This data is absent from the SOFT files, so you can't use the standard algorithms.

The only method left is to use "above X percentile of negative probes" as a cut-off for presence, which has been mentioned in a number of papersAFFX-BioB-3_at (e.g., here). While there are a handful of control probes on that chip, I'm not aware of any of them being negative probes (or if there are any, there don't seem to be enough to be useful for this purpose). Thus I don't think you can even use this method.

Of course you can still do standard differential expression, even with GEO2R in the browser.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.