Preprocessing Step In Rna-Seq Experiments

Question

[narges]

Hi,

As you know, there are some Bioconductor packages like limma, which were first introduced for microarray data but then after the emergence of RNA-seq technology, those packages can be also used for new technology by some transformations like "voom" in limma. I wanted to ask, in your opinion, what normalization or preprocessing step is needed for this kinds of packages. Or can you send me any links/papers regarding this subject. Right now I am using "voom" for this purpose but then I received negative values for the expression of the counts as it calculates the log of the transformation and I think "voom" transformation is not an optimum way for my method.

Thank you in advance

Answer 1

[Sudeep]

You can start with this or this paper to get a comparison of normalization strategies/differential expression analysis methods for read count data, plus if you go through forums like seqanswers or rna-seq blog you will find a lot of threads with discussion on this top. Here is one to begin with. "I think "voom" transformation is not an optimum way for my method" did you arrive at this conclusion or is it more of gut feeling ? :)

[narges]

Actually it is not as a result of feeling. I have done the DE analyses with other methods like DESeq and edgeR. The number of detected genes are really higher than what I got by doing voom transformation and using my own method of analysis. So I think maybe the problem is with the way I am transforming the data because I am quite sure the method is performing well regarding the microarray.

[Sudeep]

I think the statistical models behind these packages also have an effect on the number of DEGs, not just the transformation itself.