EdgeR model matrix interpretation

[sindrle]

Hi!
The EdgeR manual is quite nice, but I dont quite understand what the model.matrix for EdgeR actually will test!

status <- factor(c(rep("Diabetic",22), rep("Healthy",26)), levels=c("Diabetic","Healthy"));
patients <- factor(c(04,21,53,55,61,62,67,73,76,77,79,04,21,53,55,61,62,67,73,76,77,79, 01,08,13,70,71,72,75,81,82,83,86,87,88,01,08,13,70,71,72,75,81,82,83,86,87,88));
timepoints = as.factor(c(1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,2,2,2,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,2,2,2,2,2));
targets <- data.frame(Status=status, patient=patients, timepoint=timepoints)

design <- model.matrix(~patients + timepoints*status);
colnames(design)

> colnames(design)
[1] "(Intercept)" "patients4"
[3] "patients8" "patients13"
[5] "patients21" "patients53"
[7] "patients55" "patients61"
[9] "patients62" "patients67"
[11] "patients70" "patients71"
[13] "patients72" "patients73"
[15] "patients75" "patients76"
[17] "patients77" "patients79"
[19] "patients81" "patients82"
[21] "patients83" "patients86"
[23] "patients87" "patients88"
[25] "timepoints2" "statusHealthy"
[27] "timepoints2:statusHealthy"

What I want is to look if diabetics has changed genes at timepoint 2 compared to healthy (I have done this with DESeq2, just want to check if they match).
Will: [27] "timepoints2:statusHealthy" Give me what I want?

Any input?

Thanks!

[sindrle]

Very basic question:

patients <- factor(c(88,87,86,83,82,81,75,72,71,70,13,08,01,88,87,86,83,82,81,75,72,71,70,13,08,01,79,77,76,73,67,62,61,55,53,21,04,79,77,76,73,67,62,61,55,53,21,04))
timepoints =as.factor(c(1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,2,2,2,2,2,1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,2,2,2))
time2.healthy <- as.numeric(status=="Healthy" & timepoints==2)
time2.diabetic <- as.numeric(status=="Diabetic" & timepoints==2)
design <- model.matrix(~patients+time2.healthy+time2.diabetic)


> colnames(design)
[1] "(Intercept)" "patients4" "patients8" "patients13" "patients21"
[6] "patients53" "patients55" "patients61" "patients62" "patients67"
[11] "patients70" "patients71" "patients72" "patients73" "patients75"
[16] "patients76" "patients77" "patients79" "patients81" "patients82"
[21] "patients83" "patients86" "patients87" "patients88" "time2.healthy"
[26] "time2.diabetic"

WHERE IS PATIENT 01?

[sindrle]

lrt <- glmLRT(fit, contrast=c(0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,-1,1))

Gives:

Coefficient: -1*time2.healthy 1*time2.diabetic

This will give results for:

genes for which the change from time1 to time2 is different in the diabetic patients vs the healthy patients

[dpryan]

WHERE IS PATIENT 01?

The intercept. BTW, since you've moved the discussion to the bioconductor email list, just keep it there. It's too much work for people to jump back and forth between here and there to get the context of everything!

[sindrle]

Yes, Im sorry for that.

I was thinking that since the thread was already opened, I should update for people later.