Let me start by saying I've read all posts that are similar to my question and the insights are tremendous! However, as I've been doing bioinformatics for a little over a year now, I'm still a newbie in all possible respects, and I just want a clarification.
So here's the gist. I have replicates of three conditions (each with its own Input). What I first did, last year (when I was much more primitive in bioinformatics than I am now), I called the peaks then used a IDR to determine consensus peaks between the replicates and then I used a combination of bedtools intersect -v and -wa to determine what peaks are found in one condition vs another. The problem with this is it doesn't address intensity of the binding peaks- it could be that the peaks that are found in condition A and not in b are very low intensity binding and I may be getting rid of regions that even though all conditions have, may have strong binding in one condition and weaker binding in the other condition.
So I saw that Tommy Tang had a pipeline, so I followed his advice on making a Count Table of all the conditions and then run a DESeq2 a la RNA seq. Based on the discussion from his question on biocoductor (DESeq2 for ChIP-seq differential peaks) I am now planning to use GreyListChIP to remove the cell type/condition specific peaks from the Inputs and then running the DESeq2.
I know that diffbind has been designed for this purpose, but I can't place my bam files on my computer which is required (probably should ask a separate question if it is possible to access a bam site from a remote server to do diffbind).
I plan to do DESeq2 after I made a count table adjusting for the " Grey List" I'm wondering if the parameters for the DESeq2 should be the same as RNAseq or should I run different parameters for this ChIPseq. When I tried to run a typical dds -
dds <- DESeq(dds) and I got the following warning:
estimating size factors estimating dispersions gene-wise dispersion estimates mean-dispersion relationship -- note: fitType='parametric', but the dispersion trend was not well captured by the function: y = a/x + b, and a local regression fit was automatically substituted. specify fitType='local' or 'mean' to avoid this message next time. final dispersion estimates fitting model and testing -- replacing outliers and refitting for 2 genes -- DESeq argument 'minReplicatesForReplace' = 7 -- original counts are preserved in counts(dds) estimating dispersions fitting model and testing
Again just want to make sure I'm on the right track.