Hi, my lab plans to do scATAC-seq in a non model organism. We have generated a series of preliminary bulk ATAC-Seq runs from the same biological sample in many different technical conditions (dissociation methods, buffers, incubation time with transposase etc..., ~20 conditions). I ran a pipeline to process these results which relies on Bowtie for the alignments / MACS2 for peak calling. At this point, the alignment values are not too bad, I can open a genome browser to display peak data tracks, and the peaks look mostly consistent across runs. I have various reports (FastQC alignment rates etc...). Due to the organism I don't have well annotated TSS or a list of regions to blacklist.
What criteria should I use to select the "best" experimental conditions?
- alignment rates before/after trimming
- (normalized across runs) peak height/width
- consistency between peak positions across runs
- attempt to determine TSSs from ATAC-Seq and align peaks across that (but that's a circular argument)
- FRiP scores