Hi,

I have 3 groups(A,B,C) and triplicates per each group. I'm interested in looking at genes differentially expressed in group C but not in A & B.

I'm running the following deseq2 cmds and using 'contrast' function, but I'm not sure if it's the right way to define contrast -the results do not make sense to me esp log fold change. Any help is appreciated.

Count data

S1

S2

S3

S4

S5

S6

S7

S8

S9

ENSMUSG00000000001

12248

13314

11525

16055

12472

14372

14047

9474

13866

ENSMUSG00000000028

1053

1148

1821

1146

908

1241

1024

503

1541

Meta data

condition

S1

A

S2

A

S3

A

S4

B

S5

B

S6

B

S7

C

S8

C

S9

C

DESeq2 cmds:

dds<-DESeqDataSetFromMatrix(countData=countData,colData=metaData,design=~condition)
dds <- dds[ rowSums(counts(dds)) > 1, ]

dds <- DESeq(dds, betaPrior=FALSE)
res<-results(dds,contrast=c(-0.5,-0.5,1))

res

baseMean

log2FoldChange

lfcSE

stat

pvalue

padj

ENSMUSG00000000001

12878.6725789033

6.9595564977

0.1304387977

53.3549574268

0

0

ENSMUSG00000000028

1136.957652803

5.5701754543

0.3370332896

16.5270779638

2.34226412238182E-61

5.73408118628246E-61

Many thanks!

The numeric contrast here doesn't make sense, because the first coefficient is the intercept in this model. You should use ~0 + condition instead.