I'm trying to figure out how to perform a ChIP-sew differential binding analysis on specific regions of interest (in my case a 1kb upstream region of specific transposons).
I have ChIP-seq data for various cell lines and two histone markers of interest (H3K4me3, H3K27ac), but only one replicate of each (this is data from a few years ago so it is impossible to get more replicates). I intend on using the K4me3 and K27ac as surrogate replicates because I'm looking for enrichment in BOTH these markers in the upstream regions.
I'm not looking for novel differential binding regions across the whole genome, so the typical analysis routes of DiffBind and csaw don't seem to fit my needs.
In the DiffBind manual it says it is possible to read in peaksets that are not from peak-callers (eg: specific genomic windows), is it possible to read in a 'peakset' with a BED file with the regions that I am interested in? How would I go about doing this?
What I'd like for the analysis to accomplish:
- normalize the ChIP-seq signal to the input
- filter for specific genomic regions outlined in a BED file
- analyze differential binding in those specific regions for condition1 and condition2
- identify the differentially bound sites
Any suggestions appreciated!!
Thank you!!
Maybe check deeptools SES normalization. I personally only use inputs for peaek calling since there is (to my knowledge) no robust way to include them into a differential analysis, e.g. as an interaction, due to problems with normalization as their composition is totally different from the IP.
Not sure if this is a good idea, the common statistical frameworks rely on having a large amount of regions with most being non differential to robustly apply normalization and dispersion estimation.
Not without replicates, please use google and the search function and read on why you always need replicates in experiments. There are plenty of threads advising what to do without replicates.