ChIP-nexus peak call

Question

[Sudhir Jadhao]

Hello guys,

I want to analyze ChIP-nexus data, which software will be best to call peak?

And how the data analysis different from ChIP-seq?

Thank you in advance

Answer 1

[Satyajeet Khare]

Here is a paragraph from the ChIP-nexus paper

To test the robustness of the ChIP-nexus protocol, we analyzed the correlation between replicates at bound regions. Although peak finding algorithms are not designed for ChIP-nexus data we found that MACS 20 (version 2) and Peakzilla 2 identified thousands of binding peaks in all cases.

Peak calling method is also described in the same paper.

MACS v.2.0.10 20 was run on the ChIP-nexus replicate #1 samples and the ChIP-seq samples for TBP, Dorsal, Twist and Max using the following parameters:

macs2 callpeak –g dm –keep-dup=all –call-summits

Resulting peak summits were sorted by score and a maximum of 10,000 were retained per sample.

You will also find alignment information on their sample page in GEO.

[mlweilert]

In addition, it is possible to boost your MACS2 peak call on exo/nexus data by running macs2 callpeak –g dm –keep-dup=all –call-summits --ext-size 150 --shift -75. By extending the fragment sizes and then recentering them, you are in a sense simulating ChIP-seq data. However, keep in mind that if the positive and negative strand tracks of your ChIP-nexus signal are far apart, you may obtain a bimodal distribution of reads from this, which MACS2 should be able to interpret, but it is important to be aware of.

Answer 2

[mlweilert]

In addition to the answer by Satyajeet Khare, there are two other papers published on this information: PeakXus and QNexus. Each has their own strengths and weaknesses, so it is a good idea to read the associated papers carefully.

Additionally, ChIP-nexus data provides strand-specific information. So instead of having one profile of ChIP-seq data which is aggregated reads, you will have two profiles for each ChIP-nexus peak. This means that the interplay between the positive and negative strands will help you locate where the transcription factor is bound to the DNA.

The analysis is similar to ChIP-seq, except you will be operating at a much higher resolution and need to be aware that there is greater sensitivity in binding behavior. For example, if you see a wide and "fuzzy" peak, this is not necessarily background. Rather, there is a possibility that the transcription factor is binding weakly, binding indirectly, or is located at a low-affinity motif site.