For microarray gene expression data for
5 species: sp1, sp2, sp3, ...
7 time-points: t1, t2, t3, ...
3 replicates: r1, r2, r3, ...
And the aim is to identify genes differentially expressed genes w.r.t to sp1 (assume sp1 is reference species with whom I want compare expression values).
Using limma I compared created a contrast matrix to compare sp2-sp1, sp3-sp1, sp3-sp1.
Design matrix looks like:
design <- model.matrix(~0+ targets$Organism)
colnames(design) <- c("sp1","sp2","sp3","sp4","sp5","sp6")
head(design)
sp1 sp2 sp3 sp4 sp5 sp6
1 0 1 0 0 0 0
2 0 1 0 0 0 0
3 0 1 0 0 0 0
4 0 1 0 0 0 0
5 0 1 0 0 0 0
6 0 1 0 0 0 0
tail(design)
sp1 sp2 sp3 sp4 sp5 sp6
139 0 0 0 0 0 1
140 0 0 0 0 0 1
141 0 0 0 0 0 1
142 0 0 0 0 0 1
143 0 0 0 0 0 1
144 0 0 0 0 0 1
contrast.matrix <- makeContrasts(sp2-sp1,sp3-sp1,sp4-sp1,sp5-sp1,sp6-sp1,levels=design)
contrast.matrix
Contrasts
Levels sp2 - sp1 sp3 - sp1 sp4 - sp1 sp5 - sp1 sp6 - sp1
sp1 -1 -1 -1 -1 -1
sp2 1 0 0 0 0
sp3 0 1 0 0 0
sp4 0 0 1 0 0
sp5 0 0 0 1 0
sp6 0 0 0 0 1
But for example for sp2-sp1 set, not only genes differentially expressed in sp2 but also differentially expressed in some other species e.g., sp4 is given as output.
So, I am not sure if limma package is suitable for such time-point and species specific data?
Can anyone suggest any other approach or package to identify genes differentially expressed across species?
Limma is a completely appropriate tool for this sort of experiment (in fact, it'd probably be the most commonly used tool). Just because a particular gene is differentially expressed between sp2 and sp1 doesn't mean it can't also be DE between sp4 and sp1 (or even between sp4 and sp2).
•
written
by
Devon Ryan
OK I completely agree with you that a gene can be DE in multiple species.
But would there be any means to identify or pin-point genes whose expression is conserved in sp1,sp3,sp4 but DE only for sp2?
What if I group sp1,sp2,sp4 together say into allsps and compare sp2-allsps, to get DE in sp2 against all the remaining species?
For that sort of thing it's often best to use something like WGCNA. I say "better" because I suspect there are a bunch of patterns like this that you'd like to look at and this is a convenient* method to do that.
*This does not imply that WGCNA is easy to use and/or understand/debug.
•
written
by
Devon Ryan
Log in to answer this question.
will limma work for samples without replicates?