Problem in finding NB dispersion for my HiC datasets
1
0
Entering edit mode
@manjulathimma-11250
Last seen 4.9 years ago

Hi Aaron,

 

I have two datasets of HiC experiment on , one control and one sample. I am trying to use diffHiC to find differential interaction between them.

While going through the user manual and following the instructions, I got stuck up at Modelling biological variablility.

design <- model.matrix(~factor(c("hepg2","ago1")))
> colnames(design) <- c("HepG2","Ago1")
> design
  HepG2 Ago1
1     1    1
2     1    0
attr(,"assign")
[1] 0 1
attr(,"contrasts")
attr(,"contrasts")$`factor(c("hepg2", "ago1"))`
[1] "contr.treatment"

> y <- asDGEList(data)
> y$offset <- nb.off
> y <- estimateDisp(y,design)
Warning message:
In estimateDisp.default(y = y$counts, design = design, group = group,  :
  No residual df: setting dispersion to NA
> y$common.dispersion
[1] NA
> head(nb.off)
          [,1]       [,2]
[1,] 0.5439510 -0.5439510
[2,] 0.5342466 -0.5342466
[3,] 0.5398196 -0.5398196
[4,] 0.5375195 -0.5375195
[5,] 0.5434533 -0.5434533
[6,] 0.4076305 -0.4076305

Why am I not able to see common.dispersion?

Best,

Manjula

 

 

diffhic • 865 views
ADD COMMENT
1
Entering edit mode
@james-w-macdonald-5106
Last seen 5 hours ago
United States

The short answer is that you can't estimate the dispersion if you don't have any replicates.

ADD COMMENT
0
Entering edit mode

Also see the recommendations in section 2.11 of the edgeR user's guide, on what to do when you have no replicates.

ADD REPLY
0
Entering edit mode

If dispersion cant be estimated, then how do I create fit object to take next step of 'testing for significant interactions'. Is there any work around for handling samples without replicates?

ADD REPLY
0
Entering edit mode

See points 2 and 3 of section 2.11. However, these are only stop-gap solutions; the correct fix is to perform the experiment again to obtain a set of replicates. Never mind dispersion estimation, it's just good science - the only way to know if your result is reproducible is to do it again. You'll also get more power to detect differences when you have more samples.

ADD REPLY

Login before adding your answer.

Traffic: 260 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6