Entering edit mode
@avichaiamradipsunibech-6659
Last seen 10.6 years ago
Hi Michael,
I have a dataset of five different conditions with three repeats each
and my aim is to conduct a few different pairwise comparison between
the different conditions. I followed successfully your instructions
from Wed Jun 11 15:32:11 about how to do it in DESeq 2.
However, while analyzing the results I saw some strange numbers where
big differences did not get significant Padj- when I repeated the
analysis with uploading only the relevant conditions the Padj dropped
dramatically.
One example:
* pairwise comparison from the big data set - normalized reads
of condition a: 216,241; 204,221; 229,866. condition b: 11,192;
10,840; 9,172 Padj: 0.132
* comparison in a dataset of the relevant conditions only:
normalized reads of condition a: 220,126; 205,579; 230,317. condition
b: 11,341; 10,897; 9,233 Padj: 1.74E-215
Can you think of what might have caused this problem?
I'm wondering if I should go back and analyze each comparison
separately.
The basic script I have used:
sampleFiles <- c("P11.txt","P12.txt","P13.txt","P21.txt","P22.txt","P2
3.txt","L11.txt","L12.txt","L13.txt","L21.txt","L22.txt","L23.txt","L3
1.txt","L32.txt","L33.txt")
sampleCondition<-c("a","a","a","e","e","e","ILa","ILa","ILa","ILe","IL
e","ILe","ILh","ILh","ILh")
sampleTable<-data.frame(sampleName=sampleFiles, fileName=sampleFiles,
condition=sampleCondition)
ddsHTSeq<-DESeqDataSetFromHTSeqCount(sampleTable=sampleTable,
directory=directory, design=~condition)
colData(ddsHTSeq)$condition<-factor(colData(ddsHTSeq)$condition,
levels=c("a","e","ILa","ILe","ILh"))
dds<-DESeq(ddsHTSeq)
res<-results(dds, contrast=c("condition","a","ILa"))
res<-res[order(res$padj),]
mcols(res,use.names=TRUE)
write.csv(as.data.frame(res),file="a-ILa.csv")
All the best,
Avichai
[[alternative HTML version deleted]]
