are these results sensible?

[avilella]

I am trying NucleoATAC on datasets of a lab workflow similar to ATAC-seq but not quite, and I am wondering if the results I am getting make sense. Some samples have a decent looking VMat plot, e.g.:

Some others look a bit more flat/discontinuous:

The model plots for the nice looking ones seem good to me:

Does this look good? Anything I should play around with or change?

[AliciaSchep]

NucleoATAC is pretty specifically targeted towards ATAC-seq data. In the paper, we also showed it could be used for MNase, but that required using a V-plot learned from Mnase (default NucleoATAC one is from ATAC) data and also did not use the sequence bias model for ATAC (which NucleoATAC uses by default); additionally only the cross-correlation analysis was performed for MNase, and not an occupancy calculation. For a protocol with a different enzyme, I would not expect the V-plot to look the same as for ATAC-seq. NucleoATAC uses a V-plot learned from ATAC-seq as starting point, and adjusts based on fragment size of sample; doing so only makes sense if sample is also ATAC-seq. It also looks like the modelling of the fragment size did not work at all, as it doesn't seem like your sample has a mixture of NFR and nucleosome reads, as is the case for ATAC-seq.

In general, while an approach similar to what is used in NucleoATAC may be effective with your kind of data, using NucleoATAC is probably not appropriate 🙁

[avilella]

Hi @AliciaSchep , thanks for your reply a few months ago.

I have seen a bit of activity in the codebase and I wonder if now is a good time to bring up this ticket issue again.
From your reply above, it seems like starting with MNase-like data similar to the one we have, would require quite a lot of tweaking in the training of NucleoATAC and then running it for a given dataset.
Is this something I would be able to try following instructions available somewhere?
If this is an insurmountable amount of work, would you be able to recommend me any similar tools to NucleoATAC for my MNase-like datasets that would give me a per-position score combining read depth + insert size + nucleosome positioning information?

Thanks

[AliciaSchep]

Hi @avilella I don't have any detailed instructions for applying NucleoATAC to MNase. As for alternative methods, I'd suggest checking out the method in https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5094857/ Haven't tried it out, but it seemed really interesting when I read the paper