Question concerning normalization and log2 transformation of gene expression data from micro arrays

Question

[Nicolai Skovbjerg Arildsen]

Hi,

I have a question regarding a dataset that I inherited from a previous student at my facility. It is background corrected gene expression data done on a Illumina HumanRef-8 WG-DASL v3.0 platform. I rather new at this so I have read some papers around the subject.

However, as I have only the "background" corrected data, and no access to any CEL files and no other steps have been done, I've tried to quantile normalize the data and then log2 transform, however a problem arises.

After quantile normalization I have negative values in my data set which becomes NA after the log2 transformation. How should I treat these values, or should I just log2 transform first and the quantile normalize?

(I really think I have tried to come up with a solution on these forums, so I hope that I haven't missed any obvious searches)

Best regards

Nicolai

Answer 1

[andrew.j.skelton73]

If you've got illumina data, it's more likely to be IDAT files, CEL files are typically Affy chips. I think you should be fine using the lumi pipeline for VST/RSN normalisation of your background corrected raw data, annotate, then differential expression using Limma.

You may need to cast your raw text into a an expressionSet object, or see if the lumiR function can handle plain text.