Design matrix after RUVg correction

Question

[Shaimaa Gamal]

I have 3 estimated W (factors of unwanted variation) after RUVg correction. I am trying to include W (estimated factors of unwanted variation) into the design matrix just as recommended by Prof. Davide Risso 1 but I am concerned about how this will affect the contrast matrix? They appear as 3 extra rows compared to the desired contrast matrix (to have 5 rows and 5 columns). Is there some wrong application here? W is a matrix of 3 columns (W_1, W_2, W_3)

 design=model.matrix(~0+W+dge$samples$Stage)
 colnames(design)[1] ="W_1"
 colnames(design)[2] ="W_2"
 colnames(design)[3] ="W_3"
 colnames(design)[4] ="Control"
 colnames(design)[5] ="Stage1"
 colnames(design)[6] ="Stage2"
 colnames(design)[7] ="Stage3"
 colnames(design)[8] ="Stage4"

 ###Change columns order###
 design <- design[, c(5,6,7,8,4,1,2,3)]

 v = voomWithQualityWeights(dge, design = design, plot = TRUE)

 vfit <- lmFit(v, design)

 contrast.matrix <- makeContrasts(Stage1vsControl=Stage1-Control,
 Stage2vsControl=Stage2-Control,
 Stage3vsControl=Stage3-Control,
 Stage4vsControl=Stage4-Control, 
 levels = colnames(design))

 vfit <- contrasts.fit(vfit, contrasts=contrast.matrix)

 efit <- eBayes(vfit)

 plotSA(efit,main = "Final model: Mean-variance trend")

 summary(decideTests(efit, method = "separate", adjust.method = "BH", p.value= 0.05,lfc = 1))

[Gordon Smyth]

Cross-posted to Bioconductor https://support.bioconductor.org/p/9156456/

Please be aware that posting the same question at the same time to multiple forums without saying so is considered impolite. For my part, I do not answer questions if I notice that they have been cross-posted.

You have been getting extensive help on the Bioconductor forum, including from me (the limma author) and including for the same question that you have posted here.