I've been trying a package called 'MEDIPS' for MeDIP-seq analysis for few days now. The manual says : The genome will be divided into adjacent windows of length 100nt and all further calculations (short read coverage, differential coverage between conditions etc.) will be applied to these windows, if ws =100. I've seen a paper (PMC4213696) apart from MEDIPS's original paper that researchers are using ws = 100 only.
My question is what is the criteria for selecting window size? What if I increase the window size ? I guess the values such as 'counts', 'logFC' , 'rpkm' & eventually 'p-value' will change, is I do so.
Please help me out.
Thank you :)
The author has answered on Bioconductor:
you can certainly try other window sizes such es 300 or 500nt. This will have a strong impact on the number of significant cases after correction for multiple testing as you are massively reducing the number of windows tested for differential enrichment. Due to the same reason you should also make use of the minRowSum parameter (>=10) or focus on regions of interest such as previously calculated peak regions. For an improved MeDIP normalisation, please also consider our QSEA package.