diffbind finding not differential site
1
0
Entering edit mode
woongjaej • 0
@woongjaej-13388
Last seen 9 months ago
Korea/Seoul/Yonsei University, College …

Hi,

I'm using Diffbind to analysis differential binding H3K27Ac histone ChIP-seq site!

I'm curious whether I can find all the relative site like not differentially binded site in both conditions, up binding in condition1 and, up binding in condition2.

I gave the fold value 0 in dba.analyze, but the result give me just about | 0.3 fold | differentially binded sites.

Any suggestion will be very grateful!.

Thanks,

Woongjae

3
Entering edit mode
Rory Stark ★ 4.7k
@rory-stark-5741
Last seen 12 hours ago
CRUK, Cambridge, UK

The null hypothesis in these tests is that the site did not change binding affinity. Low p-value/FDR value indicate confidence that this is no true -- that there is a change. Higher p-value/FDR values,on the other hand, do not indicate confidence that the sites did not change, only that there is not sufficient evidence to be confident that a change occurred. Bottom line: this tool can not be used to test for unchanged site.

If you want to see sites with low fold-changes, you should set th=1 in the call to dba.report() to retrieve all the binding sites. Then you can filter by low Fold changes and higher FDR. You can then sort to see the lowest fold changes first. Here's an example:

data(tamoxifen_analysis)
rep <- dba.report(tamoxifen,th=1)
filter <- abs(rep$Fold)<1 & rep$FDR>.1
lowfold <- rep[filter,]
lowfold[order(abs(lowfold\$Fold),decreasing=FALSE),]

0
Entering edit mode

Thanks Rory!! This is very helpful solution!

Woongjae