DiffBind for Allele-specific binding analysis
2
0
Entering edit mode
Vivek.b ▴ 100
@vivekb-7661
Last seen 17 months ago
Germany

Hi Everyone

I have ChIP-Seq in a strain with defined maternal/paternal background and Now I want to use DiffBind to perform allele-specifc binding analysis. But what should be the design of analysis in this case? I can think of different ways:

  • I can compare maternal ChIP vs maternal Input (group 1) to paternal ChIP vs paternal Input (group 2), In this case I will have to call peaks separately for mat/pat allele. But I am afraid that there are not enough reads to call peaks separately on alleles. 
  • I can compare between maternal-ChIP vs paternal-ChIP (group1) and maternal Input vs paternal Input (group 2), again by calling peaks separately for ChIP and Input. 
  • I can do the peak calling first on overall ChIP vs Input. Then I count the maternal/paternal reads on the peaks on both ChIP and Input samples, do differential binding using design 1 or design 2 above.

There could be other ways as well. But I would like to have some suggestions.

 

 

diffbind chipseq • 852 views
ADD COMMENT
1
Entering edit mode
Rory Stark ★ 4.1k
@rory-stark-5741
Last seen 15 days ago
CRUK, Cambridge, UK

Some more info might help in figuring out what to do. How many separate ChIPs do you have? Do you have separate ChIPs on the maternal/paternal strands, or are you separating he maternal/paternal reads after the sequencing? How many replicates do you have of each group? How many ready for each ChIP?

In general, it should be OK to call peaks using all the reads for each ChIP, (option 3). Then you can count the reads-in-peaks separately for each condition and compare them in DiffBind. 

Cheers-

Rory

ADD COMMENT
0
Entering edit mode

Thanks Rory for your reply. In short I have ChIPs for 5 different TFs with 2 replicates each , with matching input. The ChIPs are done on the F1 hybrid cells, and I create maternal/paternal genome using SNP information. Then I map the reads into these genomes separately, I convert the coordinates back to reference genome to make the comparable, then I count maternal/paternal reads (since I have mapping flags to separate them). 

So I for each ChIP I seperate maternal/paternal reads for Control(Input) and Test(ChIP) sample. The diffbind design problem starts here.

 

ADD REPLY
1
Entering edit mode
Rory Stark ★ 4.1k
@rory-stark-5741
Last seen 15 days ago
CRUK, Cambridge, UK

Hi Vivek-

in this case, I would call peaks for the 10 ChIPs (5TF x 2 reps), and generate two .bam, files for each ChIP (maternal/paternal). in DiffBind, the sample sheet would have 20 samples, with each ChIP having the same peak file but a different bam file. Then compare the maternal to paternal (for each TF I assume).

Cheers-

Rory

ADD COMMENT
0
Entering edit mode

Can you suggest where shall I include the control (Input) files in this design? Although I am calling peaks using input first, I still expect that even on some of these peaks, input can show maternal/paternal bias. These sites should be filtered/normalized somehow since a bias in input tells me that the bias in peaks are not reliable. :-/

ADD REPLY
0
Entering edit mode

That won;t be very easy to do! One simple this would be to generate separate Input bam files (maternal/paternal), and these will be subtracted form the peak counts (depending on what scoring method you use). This will address a bit of the issues but only really clear cases. In order to really get at Input bias, you'd need to sequence deep enough to get clear deeper coverage for each version of the chromosome and call peaks separately.

-R

ADD REPLY
0
Entering edit mode

I see. Yah, I think I can just remove those samples which show input bias, in this case. So basically I can do either of the two:

1) Call peaks. Do diffbind (maternal vs paternal allele) on ChIP and Input, then remove those peaks which are biased in input too.

2) If the coverage in Input is enough and mostly comparable for maternal and paternal genome, then call peaks for each genome separately and do diffbind afterwards.

2nd one seems more neat. I wish I find enough coverage. Thanks.

ADD REPLY

Login before adding your answer.

Traffic: 298 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