[Feature Request] ATAC-seq chromatin accessibility metanode set

[BDBGenomics]

I noticed there's no dedicated ATAC-seq workflow support yet surprising given how common it's become in epigenomics studies. The platform covers a lot of transcriptomics ground but chromatin accessibility feels underserved.
At minimum, a useful metanode chain would look like:

MACS3 peak calling → TSS enrichment → FRiP scoring → differential accessibility (DESeq2/edgeR)

I work on ENCODE-compliant ATAC-seq pipelines and would be happy to implement these if the maintainers can point me to a good reference metanode to model the TypeScript spec from. Already familiar with the contribution guide.
Happy to discuss scope before opening a PR.

[u8sand]

Thanks for your interest!

Prefacing that I'm not that familiar with ATAC-seq--I'm imagining that the workflow would be not unlike the transcriptomics or metabolomics analyses which essentially:

1. turn raw data readouts into some kind of matrix (peak calling perhaps?)
2. normalization/QC/visualization (TSS/FRiP -- i'm guessing)
3. differential analysis on matrix => ranked significant differences
4. visualization of that ^
5. connect those significant differences to insights (enrichment analysis?)

You could start with the matrix --
https://github.com/MaayanLab/Playbook-Workflow-Builder/tree/main/components/data/anndata

the way it was done with metabolomics
https://github.com/MaayanLab/Playbook-Workflow-Builder/tree/main/components/data/metanndata

You can see how you derived from these matrices as in:
https://github.com/MaayanLab/Playbook-Workflow-Builder/tree/main/components/data/log_normalization

As we add new domains it seems we should improve the folder hierarchy which is currently just called data at the moment but i suppose should be called transcriptomics/metabolomics/etc.

Can you help me understand what the computational requirements for this type of analysis that might be expected? What's the typical data size, how long does it usually take to execute, etc..

[BDBGenomics]

Hey, thanks for the detailed reply. your understanding of the workflow is pretty much exactly right.

ATAC-seq ends up looking very similar to transcriptomics structurally:

1. FASTQ → alignment/peak calling → consensus peak set → peaks × samples matrix
2. QC + normalization (TSS enrichment and FRiP are mainly pre-matrix QC checks)
3. Differential accessibility on the matrix (DESeq2/edgeR)
4. Visualization/ranking of significant peaks
5. Motif/pathway enrichment downstream

AnnData actually fits really naturally here since tools like SnapATAC2 and EpiScanpy already use it heavily.

From the PWB side, the expensive parts are mostly upstream (alignment + MACS3 peak calling). Once you’re at the matrix stage, the workflow is pretty lightweight:

• usually ~50k–150k peaks across a few to a few dozen samples
• sparse AnnData objects are often only ~50–200 MB
  -DESeq2 typically runs in a few minutes on a standard machine
• no GPU or HPC needed for the normal workflow

I also agree the folder structure will probably need to evolve as more omics domains get added. Something like "data/epigenomics/atac_anndata" would make sense alongside transcriptomics/metabolomics.

I can start by scaffolding the AnnData matrix component based on metanndata if that seems like the right starting point. Happy to open a draft PR first so we can align on structure before I build things out further.