Question: Confusion over what is reference for design matrix in limma
1
gravatar for danielle.newby
2.6 years ago by
danielle.newby30 wrote:

Hi, I need some confirmation to check I am getting this right as i have been reading examples and it is just adding to the confusion!

I am doing a case versus control study for differential expression analysis using Limma package in R. My control group is "0" and my case group is "1". I am wondering what is my reference for comparison between case and control? So when i look at the output topTable FC values how do i know what is under or over expressed?

Here is my class label:

> class
 [1] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0
Levels: 0 1

So the above would imply to me that i am making the comparison between control (0) and case (1). So if i consider the first entry in my top table:

> topTable(resultmulti1,coef = 2, n = 1)

     logFC  AveExpr        t      P.Value    adj.P.Val        B
59285 3.503369 1.645189 87.00804 1.614614e-32 2.340221e-28 63.84269

Where the FC is 3.50033 is this fold change with reference to the control? The reason i ask is that if i look at the mean expression for this gene for the two class labels:

> typeMean <- tapply(expression_iGiS[genename,], phenodata$class, mean)

> typeMean

         0          1 
-0.4548455  3.0452115

This would imply that my class 1 is the reference as if you take away 0-1 this gives - 3.50033 but if you do 1-0 this gives you +3.50033 (which is in the fold change column.

The only thing i can think of is that my design matrix is influencing what is my reference: The deisgn matrix is made:

>phenodata <- as.data.frame(cbind(c(rep(1, 15),rep(0 ,10))

>colnames(phenodata) <- c("class")

>design <- model.matrix(  ~ class  )

This create a matrix which includes an intercept (which is all 1's) and my class variable. Is the intercept being set as 1 effecting my model and what is being used as a reference?

Any comments would be really helpful! I want to know if i can say for example gene 59285 is upregulated in control compared to case. From the lack of clarity in examples i am unsure if this is the right away around. The resource i used was:http://kasperdanielhansen.github.io/genbioconductor/html/limma.html

Thanks Danielle

limma designmatrix R • 1.2k views
ADD COMMENTlink modified 2.6 years ago by Devon Ryan89k • written 2.6 years ago by danielle.newby30
1
gravatar for Devon Ryan
2.6 years ago by
Devon Ryan89k
Freiburg, Germany
Devon Ryan89k wrote:

0 is the reference, since it's lexicographically first (and given first in levels(class)). This also makes sense given that the example fold-change is basically mean(1) - mean(0).

ADD COMMENTlink written 2.6 years ago by Devon Ryan89k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1797 users visited in the last hour