Entering edit mode
Giulio Di Giovanni
▴
540
@giulio-di-giovanni-950
Last seen 10.2 years ago
Hi, I'm analyzing a dataset of 50 samples. Each of these samples has a
dye-swapped replicate. All the replicates are side by side in the
dataset (1, -1, 2, -2, 3, -3, etc...)
I'm interested in some contrasts between predefined groups.
I built a design matrix with the 1 and -1 that correspond to the
samples (without or with dye-swap) belonging to a each group (sick1,
sick2, sick3):
sick1 sick2 sick3
1 0 0
-1 0 0
0 1 0
0 -1 0
1 0 0
-1 0 0
0 0 1
0 0 -1
0 0 1
0 0 -1
...
..
But this is not enough to proceed the analysis, because the 1s and -1s
are replicates.
So I tried to estimate the correlation within replication block:
biolrep <- c(1,1,2,2,3,3,..... 50,50)
corfit <- duplicateCorrelation(data, design, ndups = 1, block =
biolrep)
And I have
corfit$consensus = -0.84
(good, they are dye swapped)
And I fit the model
fit <- lmFit(data, design, block = biolrep, cor = corfit$consensus)
Then I create and extract my contrasts:
cont.matrix <- makeContrasts( sick1.2 = sick1-sick2 ...etc)
fit2 <- contrasts.fit(fit, cont.matrix)
fit2 <- eBayes(fit2)
I'm not sure I understood very well everything. There's some mistakes
I'm making?
Any suggestion will be really appreciated...
Thanks
Giulio
_________________________________________________________________
[[alternative HTML version deleted]]