Differentially expressed genes analysis Limma microarray

Question

[judhenaosa]

Hi.

I have a question. I am working with a microarray with the next samples:

• control_rep_1
• control_rep_2
• virusA_rep_1
• virusA_rep_2
• virusA_rep_3
• virusB_rep_1
• virusB_rep_2
• virusB_rep_3
• virusA_virusB_rep_1
• virusA_virusB_rep_2
• virusA_virusB_rep_3

I have to extract the differentially expressed genes:

• control vs virusA
• control vs virusB
• control vs virusA_virus_B
• virusA vs virusB

I have to do so using Limma. In the case of the virusA, I am using the next design before to run the lmFit function:

virusA.design <- data.frame(REF = 1 , ctrl_vs_virusA = c(0,0,1,1,1,-1,-1,-1,-1,-1,-1))

virusA <- lmFit(normalised_data, virusA.design)

virusA <- eBayes(virusA)

But the adjusted p-values make no sense. I guess I am doing something wrong.

Do you have an idea about how should I design the different comparisons?

Thanks in advance,

Juan

Answer 1

[Devon Ryan]

That's a very confusing way of going about things and I expect some of the resulting coefficients will be twice what they should. Please make a design as follows:

metadata = data.frame(groups=c("control", "control", rep("virusA", 3), rep("virusB", 3), rep("virusAB", 3)))
design = model.matrix(~groups, metadata)

That will be much simpler.

[judhenaosa]

Thanks for your help.

My next question is about the lmFit function. I mean, I want to see the specific effect of the control samples over the treatment samples and the effect of the virusA vs virusB, and extract the differentially expressed genes in each individual case. However, I do not know how to do specify this in the script.

Thanks in advance,

Juan

[Devon Ryan]

Please have a look at the limma vignette, there are examples of using both contrasts and extracting coefficients.