This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Query for limma and DEseq2

Hi

I want to show the common differentially expressed genes from two data set.One is affymatrix data and the other is RNAseq dataset from GEO.For that atfirst i have to analyze the dataset individually.So will i can analyze affymatrix data with limma and the RNAseq data with DEseq2.I want to know ,will this logical or I have to analyze both with the same package?

deseq2 limma
• 3,272 views • link
updated by Ram Ram • written by fahim • fahim

So according to your reply i can use limma for affymatrix data and DEseq2 for RNAseq data.Is that clear sir.I want to add something in this .The affymatrix data that i choosed,contain GEO2R option,website generated.So in there limma package is used.And they count the signficant genes based on adj.p.value.So if i analyze my RNAseq data with limma,with coding sign gene by adj.p.value and log2fc,I found zero sign gene.But if I analyze my RNAseq data with DEseq2 i can find my sign gene with coding adj.p.value and log2fc. So this is the main confusion occured.

• 0 views • link
written by fahim • fahim

2 answers

Depends a lot on what kinds of questions you are trying to answer with these datasets, what kind of metadata you have for each dataset, and also the limitations when trying to compare datasets from two different technologies.

Also, if you know both tools well, you can always run limma and DESeq2 on the RNA-seq and compare how different these two sets of results are prior to the affymatrix comparisons. Many times there is good overlap between limma and DESeq2, but each tool has a pro depending on the design of the study (e.g.: if you need to include a random factor limma is more suitable).

• 0 views • link
written by lianov • lianov

To clarify what lianov says, traditionally, we would use just limma for processing microarray data, and then edgeR or DESeq2 for processing [bulk] RNA-seq.

limma can also be used for RNA-seq data via the edgeR-voom-limma ('limma-voom') pipeline, examples of which can be found here:

In your particular case, please use limma for processing the Affymetrix data, and DESeq2 for the RNA-seq. You can then conduct a meta-analysis on the derived test statistics.

• 0 views • link
written by Kevin Blighe Kevin Blighe

Kevin, I am unclear what objective evidence your advice is based on. limma-voom has consistently been ranked as at least as good as specialist RNA-seq tools in third party comparisons. I wasn't going to weigh in here, but your advice does require a response. Why would OP not use the same package for both analyses if they choose? In seems an obvious simplification.

• 0 views • link
written by Gordon Smyth Gordon Smyth

Yes, agree with Gordon here. I truly appreciate a number of tools for this type of analysis including DESeq2, EdgeR and limma and why I say in certain cases one may be better than another but we need to know more about the design to suggest that. Overall, under simple comparisons for example, I have seen a large overlap in the results between DESeq2 and limma-voom.

For reference/anyone who may not be aware of limma being used in RNA-seq, here is a great sample workflow (by Gordon and colleagues): https://f1000research.com/articles/5-1408

• 0 views • link
written by lianov • lianov

The OP can choose whichever package(s) they wish.

Scenario 1

  • Microarray: limma
  • RNA-seq: limma-voom

Scenario 2

  • Microarray: limma
  • RNA-seq: DESeq2

There is neither anything better nor worse than anything else. We could produce study after study that tips the odds in favour for one package or another, but then we'd have wasted years of our lives

• 0 views • link
written by Kevin Blighe Kevin Blighe

Of course, I absolutely love limma and regard it as one of the greatest contributions to bioinformatics

• 0 views • link
written by Kevin Blighe Kevin Blighe

So according to your reply i can use limma for affymatrix data and DEseq2 for RNAseq data.Is that clear sir.I want to add something in this .The affymatrix data that i choosed,contain GEO2R option,website generated.So in there limma package is used.And they count the signficant genes based on adj.p.value.So if i analyze my RNAseq data with limma,with coding sign gene by adj.p.value and log2fc,I found zero sign gene.But if I analyze my RNAseq data with DEseq2 i can find my sign gene with coding adj.p.value and log2fc. So this is the main confusion occured.

• 0 views • link
written by fahim • fahim

Log in to answer this question.

More posts like this