Hi,

I am analyzing mRNA-Seq dataset using EdgeR package, I have question while extracting co-efficient comparisons after fitting the model (example below). In limma, I use simple design <- model.matrix(~0+Culture_Types) with corfit (as below), however, doing this first time time EdgeR.

corfit <- duplicateCorrelation(df, design, block=sample_ann$Pt) 
fit <- lmFit(df,design,block=sample_ann$Pt,correlation=corfit$consensus)

Here is the example of the sample annotations and design:

dput(xx)
structure(list(Samples = c("Sample_1", "Sample_2", "Sample_3", 
                           "Sample_4", "Sample_5", "Sample_6", "Sample_7", "Sample_8", "Sample_9", 
                           "Sample_10", "Sample_11", "Sample_12", "Sample_13", "Sample_14", 
                           "Sample_15", "Sample_16", "Sample_17", "Sample_18"), Patient_ID_v1 = c("S13", 
                                                                                                  "S13", "S13", "S18", "S18", "S18", "S21", "S21", "S21", "S47", 
                                                                                                  "S47", "S47", "S61", "S61", "S61", "S70", "S70", "S70"), Culture_Types = c("CDxx", 
                                                                                                                                                                             "CDvv", "CDzz", "CDxx", "CDvv", "CDzz", "CDxx", "CDvv", "CDzz", 
                                                                                                                                                                             "CDxx", "CDvv", "CDzz", "CDxx", "CDvv", "CDzz", "CDxx", "CDvv", 
                                                                                                                                                                             "CDzz"), Cohorts = c("A", "A", "A", "A", "A", "A", "B", "B", 
                                                                                                                                                                                                  "B", "B", "B", "B", "A", "A", "A", "B", "B", "B")), class = "data.frame", row.names = c(NA, 
                                                                                                                                                                                                                                                                                          -18L))
#>      Samples Patient_ID_v1 Culture_Types Cohorts
#> 1   Sample_1           S13          CDxx       A
#> 2   Sample_2           S13          CDvv       A
#> 3   Sample_3           S13          CDzz       A
#> 4   Sample_4           S18          CDxx       A
#> 5   Sample_5           S18          CDvv       A
#> 6   Sample_6           S18          CDzz       A
#> 7   Sample_7           S21          CDxx       B
#> 8   Sample_8           S21          CDvv       B
#> 9   Sample_9           S21          CDzz       B
#> 10 Sample_10           S47          CDxx       B
#> 11 Sample_11           S47          CDvv       B
#> 12 Sample_12           S47          CDzz       B
#> 13 Sample_13           S61          CDxx       A
#> 14 Sample_14           S61          CDvv       A
#> 15 Sample_15           S61          CDzz       A
#> 16 Sample_16           S70          CDxx       B
#> 17 Sample_17           S70          CDvv       B
#> 18 Sample_18           S70          CDzz       B

Patient_ID_v1 <- factor(xx$Patient_ID_v1)
Culture_Types <- factor(xx$Culture_Types)
design <- model.matrix(~0+Patient_ID_v1+Culture_Types)
## dispersion estimated:
y <- estimateDisp(y,design)
fit <- glmQLFit(y, design)

## What are the co-efficient to be chosen here, `coef = `

1. To detect genes that are differentially expressed in CDxx vs CDzz ?:

qlf <- glmQLFTest(fit, coef=????)

2. To detect genes that are differentially expressed in CDxx vs CDvv ? :

qlf <- glmQLFTest(fit, coef=????)

3. To detect genes that are differentially expressed in CDzz vs CDvv ?:

qlf <- glmQLFTest(fit, coef=????)

OR,

Maybe a one table to all comparisons of interest like we do in contrast.matrix:

con <- makeContrasts(
  CDxx_vs_CDzz = CDxx-CDzz,
  CDxx_vs_CDvv = CDxx-CDvv,
  CDzz_vs_CDvv = CDzz-CDvv, levels=design)

Thank you,

Mohammed

Specifying coefficients and contrasts is almost exaclty the same in edgeR as in limma so, if you know limma, then use the same rules in edgeR.