I think signal generation step is not belong to macs2, since it is
refer to mapping result, mapping let us know generated signal.
And peak calling is a function of macs2 but not mark here, actually,
I can't tell the difference between peak calling and signal
generation? For this fig, they need same input, bam or sam, right?
Which step's result can be as input to IGV? I used fold change
over control bigwig files as input, worked, but some slices says after peak calling?
I just have fold change over control and signal p-value result
bigwig files now, so if I need to do peak calling, I must get
mapping results bam or sam? Can't transfer bigwig files as input, right？ I
mean information equivalence is not problem? Some tips, thanks！
You're right that the "signal generation" part is confusing there. MACS2 can output an optional bedGraph track (the --bdg option I think) that is essentially a coverage/signal track. It's highly likely that that's what's intended there. Truth be told I find bigWigs from bamCoverage to be more useful, but I'm rather biased.
Peak calling and signal generation are part of the same step, it's unfortunate that they show them to be unrelated, since you'd just run MACS2 once.
You can use the bigBed files and the bedGraph (N.B., you would be better served with a bigWig file, either by converting the bedGraph to bigWig or simply using bamCoverage from deepTools).
While it's possible to do peak calling on a bedGraph file with MACS2, I don't generally recommend doing it unless you really really know what you're doing. MACS2 callpeak essentially makes a bedGraph file internally, but you'd need to match the normalization it's doing and then run macs2 bdgpeakcall. I strongly encourage you to try and track down a BAM file instead.