Question

Anova table edgeR limma

0

Entering edit mode

KATHERINE CHUERE NUNES • 0

@2fbbe51a

Last seen 18 months ago

Brazil

Hello everyone!

I am trying to understand this output table:

Example:

Coefficient: LR test on 3 degrees of freedom
          ID logFC.AvsB logFC.AvsC logFC.AvsD logFC.DvsB logFC.DvsC logFC.CvsD   logCPM        F   PValue      FDR
Transcript 1        0.5       -0.2        0.6       -0.8      -0.07        0.4 8.407242 76.38408 1.78E-10 2.96E-07

In the condition A, is the transcript 1 up or downregulated in the comparison AvsB?

Can anybody help me, please?

Thank you. Katherine

limma RNASeq edgeR • 895 views

ADD COMMENT • link 18 months ago KATHERINE CHUERE NUNES • 0

score 2 · Accepted Answer · 2023-05-12

2

Entering edit mode

Gordon Smyth 52k

@gordon-smyth

Last seen 9 hours ago

WEHI, Melbourne, Australia

To answer your question for certain we would need to see your earlier code, in particular how you defined the contrasts that have been tested.

We can guess though. The coefficient called "AvsB" is 0.5, which is positive, so that would indicate up-regulation in A vs B if you have followed the usual naming conventions. The F-test and p-value are for all comparisons, of course, so one cannot say that the AvsB comparison is statistically significant on its own.

ADD COMMENT • link 18 months ago Gordon Smyth 52k

0

Entering edit mode

Hello Gordon Gordon Smyth !

The code was:

con <- makeContrasts(
AvsB = A - B,
AvsC = A - C,
AvsD = A - D,
BvsC = B - C,
BvsD = B - D,
CvsD = C - D, levels=design)

anov<-glmQLFTest(fit,contrast=con)
topTags(anov)

So in "AvsB", the transcript is up-regulated in the group A? And in the group B, down-regulated. Correct?

I thought it would be like in paired comparison ''coeficient'', the up or downregulated is seen on the second group (for example, in B):

qlf.BvsA<-glmQLFTest(fit,contrast=my.contrasts[,"BvsA"])                
topTags(qlf.BvsA)

Coefficient: 1A 1 B

logFC logCPM F PValue FDR

Transcript 1 0.5200925 13.4856 29.69291 1.58E-05 0.001140845

Best wishes, Kathy

ADD REPLY • link 18 months ago KATHERINE CHUERE NUNES • 0

1

Entering edit mode

You have computed A - B and got a positive result (0.52) so the expression must be higher in A than in B.

The edgeR contrasts mean exactly what they appear to mean. They just follow basic arithmetic. There's nothing hidden or tricky.

I thought it would be like in paired comparison ''coeficient'', the up or downregulated is seen on the second group (for example, in B):

Both topTags tables give the same coeffcient for AvsB with the same meaning in each case.

ADD REPLY • link 18 months ago Gordon Smyth 52k

0

Entering edit mode

Thank you Gordon Gordon Smyth ! You helped me a lot!

Best wishes, Kathy

ADD REPLY • link 18 months ago KATHERINE CHUERE NUNES • 0