Add store_bdg, SPMR, broad parameters for calling peaks using MACS3

[beyondpie]

Hi Kai,

I add three parameters for peak calling.

1. On bed graph generation:

   • SPMR and store_bdg, I open the two parameters since we notice that MACS2/3 can generate the smoothed bedgraphs for our later deep learning model usage.

2. On broad peak calling:

   • add broad to allow broad peak calling. We typically use this for broad peak calling on some histone modifications.

But I am not sure if this is enough for the peak calling function since I notice that you actually have a rust layer for the real task. Another question I have: there is nomodel parameter in the list. According the MACS3 README, it seems to be needed for extsize parameter.

By default, I set all of them false in order to keep the default behavior of your function.

Thanks! Songpeng

[kaizhang]

Thanks for the pull request!

1. I suggest remove store_bdg, SPMR. And implement bedgraph smoothing in https://kzhang.org/SnapATAC2/api/_autosummary/snapatac2.ex.export_coverage.html. This should be easy.
2. Have you check if the broad peaks called by macs3 and snapatac2 agree with each other?

[beyondpie]

Hi Kai, Thanks for your suggestions!

1. I will give it a look. In MACS2/3, it allows us to shift and extend the fragments, not sure if your function of export_coverage support this. I think this is the main reason why we choose MACS2, otherwise, deepTools or bedtools also provide the generation of tracks with more parameters.

2. I did not use SnapATAC2 for peak calling later since in our current pipeline, we use bam files to directly call peaks. But Johnny(@zwang0715) recently used SnapATAC2 for peak calling. Maybe he can give us some comments here. BTW, is the parameter nomodel used in your peak calling pipeline?

Thanks! Songpeng

[kaizhang]

yes, nomodel is used

[beyondpie]

@kaizhang https://kzhang.org/SnapATAC2/api/_autosummary/snapatac2.tl.macs3.html From this doc, it supported the unmerged content, like save_bdg. Do you think we should merge this commit or update the document to reflect the stable version?

Thanks! Songpeng

[kaizhang]

Broad peak calling was added in c60ddecf111d651c4e24a5ef1845d6505553c485. I'm closing this pull request.