Question: coef for limma differential expression analysis
0
2.5 years ago by
Dave0
Dave0 wrote:

ArrayExpress dataset of 5 individuals 2 different tissues: "V" and "A". I want to compare the gene differential expression of "V" and "A" as paired sample for each individual.

I have used this code so far.

library(ArrayExpress)
library(limma)
rawdata=ArrayExpress("E-blablabla")
eset<-rma(rawdata)
Patients<-factor(rep(1:5,2))
Tissues<-factor(c(rep("V",5),rep("A",5)))
design<-model.matrix(~Patients+Tissues)
fit<-lmFit(eset,design)
fit<-eBayes(fit)


According to limma User's Guide i should continue with

topTable(fit,coef=??)

What should coef be? I do not understand their example in the User's Guide (coef="TreatT").

If I type

colnames(design)

I get

[1] "(Intercept)" "Patients2"   "Patients3"
[4] "Patients4"   "Patients5"   "TissuesV

Coefficient should correspond to what exacly?

Also, how do I write to file for easier interpretation?

Thanks

limma de • 523 views
modified 2.5 years ago by svlachavas690 • written 2.5 years ago by Dave0
Answer: coef for limma differential expression analysis
0
2.5 years ago by
svlachavas690
Greece/Athens/National Hellenic Research Foundation
svlachavas690 wrote:

Dear Dave,

if i understand well from your design,

you would like to conduct a paired moderated t-test. Your design above, essentially represents an additive model, with your main coefficient of interest would be the last, that is  the coef="TissuesV"

which represents the average log2 fold change of tissue V versus tissue A in all patients (as the level V of the Tissues factor variable is the second level). Most probably, the other coefficients from 'Patients2" to "Patients5", represent the log2 fold change between patient1 and every other patient present.

Best,

Efstathios