Hello everyone,

I came across some chanllangings when using mROAST to test a small gene set in my data. My sample contains two cell lines, each with a time series design. I'd like to use makeContrast to build and compare the time series responses of cell2 relative to cell1 in these gene sets, just like what I did in camera.

However, I found it is not feasible. The results I'm getting by using the following codes were completely unreliable. I don't know where should I add the makeContrast design. I think my codes only compare 0.5h vs 0h regardless of cell line factor.

I'd like to know how to achieve my purpose using mROAST.

All the best,

Jiahao

time <- "0.5h"
target.i <- targets[targets$Time == "0h"|targets$Time == time,]
sample.i <- rownames(target.i)
expr.i <- expr[,sample.i]
treatment = factor(target.i$design, level=c("cell1_0h","cell1_0.5h",
                                            "cell2_0h","cell2_0.5h"))
design <- model.matrix(~ 0 + treatment)

contrast <- makeContrasts(
  t0.5h_vs_t0h = (cell2_0.5h - cell2_0h) - (cell1_0.5h - cell1_0h),
  levels = design)

voom <- voom(expr.i, design, plot = FALSE)
roast.i <- mroast(voom, GeneList, design = design, contrast = ncol(design), nrot = 1000)

You are specifying that the last column of your design should be used as the contrast. You should instead use contrast[,1], which will use the contrast you want. From ?mroast:

contrast: contrast for which the test is required.  Can be an integer
          specifying a column of 'design', or the name of a column of
          'design', or a numeric contrast vector of length equal to the
          number of columns of 'design'.