Peak width and resizing peaks, bulk ATAC-Seq data, and other small issues

[chrarnold]

Hi, first thanks for the package, it seems to work nicely. I however have a couple of questions that I think can be clarified in the Docs or somewhere else:

1. When peaks do not have equal width, a warning is raised. How are results expected to change if peaks have a different width, how "serious" should I take this warning? I did a quick test and a) resized the peaks to a width of 500 and b) left them untouched. I am getting identical deviations, which I find a bit suspicious, is this expected?
2. Is it statistically valid in the chromVAR framework to compute the rowMeans from the deviations for each of the groups in order to find out which motives are different?
3. For bulk ATAC-Seq data, I currently set norm = TRUE for the expected counts. I guess I can ignore the warning here as sparsity is not an issue?

Thanks!

[AliciaSchep]

1. Identical deviations there does seem weird. Can you share how/when you changed the peak width? In terms of the warning, it is just a warning and not an error, because in practice it generally doesn't make a huge difference. There is a supplemental figure & discussion section in the paper devoted to the issue of peak width if you are interested in reading more.
2. For comparing groups, there is a function for comparing accessibility for a motif between groups of cells or samples -- see https://greenleaflab.github.io/chromVAR/articles/Articles/Applications.html#differential-accessibility-and-variability.
3. By warning, do you mean the recommendation in the documentation or a literal warning that occurs? If the documentation, then yes, if the sequencing depth is fairly high, then that method is reasonable. Likely a good idea if you don't have all that many samples and there is an imbalance in the sequencing depth between sample.

[ashuchawla]

Hi AliciaSchep,

I am unsure of the step where I could add norm = TRUE option in chromVar as mentioned by chrarnold here. Could you please point me to the right command?

I am working with bulk ATACSeq data and trying to combine datasets run on separate days. Is there an option for overcoming batch effects in chromVar?

Could I use other tools to remove the batch effects in read counts and input that in chromVar?

Thanks for your help,

Ashu