with or without --nomodel

[Abhishek Singh]

Hi,

I am bit puzzled and don't know the correct way.

When use option --nomodel I get 3000 peaks  and when I don't use this option I get 1200 peaks.

Here are my two commands

macs2 callpeak -t Sort.Ref.G1.bam -c Ref.All_inputs.merged.subsampled.bam -n Macs.G1  --gsize hs --nomodel

the above gives 3000 peaks

macs2 callpeak -t Sort.Ref.G1.bam -c Ref.All_inputs.merged.subsampled.bam -n Macs.G1  --gsize hs

this one gives 1200 peaks.

I want to know, which is better or the correct way to call peaks, command 1 or command 2.

Could you please let me know.

Best regards

Abhishek

[John Urban]

Is the experiment ChIP-seq of a point-source (narrow binding site) TF (or other thing) or of some widespread signal?

If it is the former, using the model might give you higher specificity. If it is the latter, it makes more sense to use --nomodel.

How do the peak sets relate to each other? 

e.g. Are the 1200 peaks all/mostly represented in the peak set with 3000 peaks?

What does the macs2 output say when when trying to find the fragment length to use with the model (i.e. when --nomodel is turned off?)...?

[Abhishek Singh]

Hi,

Thank you for your reply.

Well the ChIP seq is for TFs. 

The macs suggests to use --nomodel when it is not specified.

I have not made the overlap.

Given your inputs I was thinking if using --nomodel with default fragment length 200.unless their is,a compelling reason to not to do so.

Thank you.

--
You received this message because you are subscribed to a topic in the Google Groups "MACS announcement" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/macs-announcement/VdH2agcGRX0/unsubscribe.
To unsubscribe from this group and all its topics, send an email to macs-announcement+unsubscribe@googlegroups.com.
To post to this group, send email to macs-announcement@googlegroups.com.
Visit this group at https://groups.google.com/group/macs-announcement.
For more options, visit https://groups.google.com/d/optout.

[John Urban]

Sounds like a good idea (assuming your fragment lengths were ~200 bp).

It is also good practice to run the data through at least one other peak caller.

[Abhishek Singh]

Yes, I agree that an overlap from another peak caller will be more helpful.

On Thu, Feb 15, 2018 at 6:19 PM, John Urban <mr.joh...@gmail.com> wrote:

Sounds like a good idea (assuming your fragment lengths were ~200 bp).

It is also good practice to run the data through at least one other peak caller.

--