Selecting appropriate bin size

[genesiG]

I have tested two different bin sizes and got very different scaling factors.

The ChIP-seq dataset is from a paper that treated cells with PRC2 inhibitor. The antibody targets H3K27me3.

Sample metadata looks like this:

ID ANTIBODY GROUP
G477-UNC1999_H3K27me3.final.sorted.bam H3K27me3 UNC1999
G477-DMSO_H3K27me3.final.sorted.bam H3K27me3 DMSO

When using 1kb-sized bins, the scaling factors look like:

     H3K27me3.UNC1999 ,     ave.SF = 1.07
     H3K27me3.DMSO ,        ave.SF = 1
    [--done--]`

When using 10kb-sized bins, the scale factors are much different:

     H3K27me3.UNC1999 ,     ave.SF = 4.6
     H3K27me3.DMSO ,        ave.SF = 1
    [--done--]``

The original paper does not contain a WB showing the reduction in H3K27me3 upon treatment, only spike-in normalized bigwig tracks for the samples

I was wondering how would I go about selecting the bin size for the studies and how they impact the results for broad marks like H3K27me3.

Thanks!
@hongjianjin

[hongjianjin]

The average size of H3K27me3 peaks is approximately 2kb. Usually we would expect bin size is smaller than that (not larger) to achieve better resolution or biologically meaningful result. Scaling factor is about stable using bin sizes range between 200bp and 2000bb.