IDR analysis on 100 different peak files

[kiran girdhar]

Hi

I have a 100 peak files called using MACS2 at pvalue =0.01 and I am wondering 

1) what is the best way to determine the one pvalue threshold for all 100 peak files to filter out the signal.

I do not have replicates for 100 bam files

the number of peaks I am getting ranges from 70K-150K 

[Anshul Kundaje]

What type of ChIP-seq data is this? TFs or histones?

--
You received this message because you are subscribed to the Google Groups "idr-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to idr-discuss+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[kiran girdhar]

histone modification

On Tuesday, 25 July 2017 18:15:09 UTC-4, Anshul Kundaje wrote:

What type of ChIP-seq data is this? TFs or histones?

On Tue, Jul 25, 2017 at 3:13 PM, kiran girdhar <kiran.gi...@gmail.com> wrote:

Hi

I have a 100 peak files called using MACS2 at pvalue =0.01 and I am wondering 

1) what is the best way to determine the one pvalue threshold for all 100 peak files to filter out the signal.

I do not have replicates for 100 bam files

the number of peaks I am getting ranges from 70K-150K 

--
You received this message because you are subscribed to the Google Groups "idr-discuss" group.

To unsubscribe from this group and stop receiving emails from it, send an email to idr-discuss...@googlegroups.com.

[Anshul Kundaje]

Can you provide more details on which precise modifications. It matters depending on whether they have narrow or broad regions of enrichment.

-Anshul.

To unsubscribe from this group and stop receiving emails from it, send an email to idr-discuss+unsubscribe@googlegroups.com.

[kiran girdhar]

sure

I am working on H3K4me3 (100 files) and H3K27ac(100 files) datasets

corresponding to these HM I have narrow and broad peak regions respectively.

This is the datasets on humans.

[Anshul Kundaje]

You can just use our histone pipeline here https://github.com/kundajelab/chipseq_pipeline . We have a protocol for single replicate analysis that works well in practice.

To unsubscribe from this group and stop receiving emails from it, send an email to idr-discuss+unsubscribe@googlegroups.com.

[kiran girdhar]

which means running IDR on peak sets from self-replicates of each sample. right?

[Anshul Kundaje]

Correct.

-A

To unsubscribe from this group and stop receiving emails from it, send an email to idr-discuss+unsubscribe@googlegroups.com.

[Anshul Kundaje]

Well actually correction - We dont use IDR for the histone pipeline as it is gets too stringent with MACS2's scoring and ranking which has several strange artifacts. Instead of just perform a simple overlap of peaks that are present in both pseudoreplicates.

-Anshul.

[kiran girdhar]

 We dont use IDR for the histone pipeline as it is gets too stringent with MACS2's scoring and ranking which has several strange artifacts.

Can you share any example from ENCODE/ROADMAP dataset, where you have observed these artifacts? I am curious or interested to see these artifacts on MACS2 peaks 

[Anshul Kundaje]

You can try it on your data itself with IDR and with just naive overlap. The rank-rank IDR plots will show you the issues with the ranks. More so if you activate gapped or broad peaks in MACS2 the scoring function is totally messed up so the scores and ranks are non-sensical. This has been communicated to the developers and we are working with them on a fix.

-Anshul.

To unsubscribe from this group and stop receiving emails from it, send an email to idr-discuss+unsubscribe@googlegroups.com.