macs2 for replicates analysis

[Minghui Wang]

Hi Tao,

    I am using macs2 for histone marks analysis, and my experiments considered whether this is some different between different time points. Our experiments has 2 time points, and each time has three replicates. I want to analyze it using strategy similar as RNA-seq analysis, but I need to define the unite first (for example, Gene-base or transcripts-base). In our analysis, I want to define histone binding enrichment region as a unite, but the replicates give different peak region, most of them overlapped with each other when considering 50% overlapped. I hope you can give me some suggestions about how to analyze it to define enrichment region using macs2, which used for further identifying temporal differences between them.

    Best,

    Minghui Wang

  

[Davide Cittaro]

My 2 cents:
use macs to call peaks on each sample, then try DiffBind feeding it with original alignments and the "universe" peakset obtained by merging all peaks you could call with macs.
AFAIK macs cannot handle replicates natively.

d

> --
> You received this message because you are subscribed to the Google Groups "MACS announcement" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to macs-announcem...@googlegroups.com.
> To post to this group, send email to macs-ann...@googlegroups.com.
> Visit this group at http://groups.google.com/group/macs-announcement?hl=en.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

/*
Davide Cittaro, PhD

Coordinator of Bioinformatics Core
Center for Translational Genomics and Bioinformatics
Ospedale San Raffaele
Via Olgettina 58
20132 Milano
Italy

Office: +39 02 26439211
Mail: cittaro...@hsr.it
Skype: daweonline
*/

[Tao Liu]

Right. MACS doesn't consider replicate consistency. What you could do, is to first calculate genome-wide signal correlation between replicates (e.g. wigCorrelate from UCSC on the bigwig files converted from bedGraph from MACS). Then if after you get rid of bad replicate and only keep good ones, you can run MACS on each replicate then try IDR method <https://sites.google.com/site/anshulkundaje/projects/idr> to get consensus peaks.

-T