Hello,
I'm working on a simple bulk RNA-seq project from brain tissue and would like to use WGCNA downstream of DEG analysis with DESeq2. I'm using the TxImport to DESeq2 approach.
The documentation nicely illustrates how to use VST or rlog to get normalized data for downstream analyses, and involving covariates in the DEG analysis is also nicely integrated. However, I cannot figure out how to get my normalized values to be weighted by the covariates matrices or even how to extract the covariate matrix to get this weighted table myself. I'm specifically concerned with integrating one continuous covariate (RIN) and two categorical covariate (Gender and Batch). I apologize if this has already been answered, but I'm having a hard time finding the solution. Please respond if anyone knows how to do this. Thanks!
Do you want to regress-out the effect of those? Check
removeBatchEffect
from limma orCombat-Seq
from sva package.I do want to regress out the effects. I've been having trouble getting ComBat to work, but I'm going to give it another shot with the VST data. Unfortunately, ComBat only seems to be able to deal with categorical covariates, so I was hoping there was a better way as the majority of my noise is coming from RIN (continuous). If it comes to it, I'll switch over to limma but I was hoping there was a nice solution through DESeq2.
DESeq2
does not help here since it never manipulates the normalized counts directly but includes them into its GLM.Combat-Seq
is pretty simple for the end user, it operates on the raw counts, and these you can then put intovst
, or simply limma on the vst directly. Both support categorical and continuous variables fwik. I personally useCombat-Seq
these days simply because it does not create these annoying negative values for lowly expressed genes with expression close to zero.Thanks for the replies. I went with Limma because it was suggested in the DESeq2 manual and because I've had issues with getting ComBat to work in the past. Plus, my version of R or RStudio (I can never tell which when things don't install correctly) doesn't seem to support Combat-Seq. At any rate, I think it worked, but if you (or anyone else in internet land) can check to see if it's done correctly, that'd be huge. (NOTE: for sake of space, I'm only showing 8 of 44 samples)
Is that about right? Again, thanks for all the help!
Guess that is ok, see also https://support.bioconductor.org/p/85202/ to maybe have it all a bit "cleaner".
Excellent! Thanks so much!