search

buenrostrolab / stimATAC_analyses_code

All code associated with manuscript detailing scATAC and scRNA-seq following stimulus of PBMCs
37 stars 12 forks source link

BuenRTools package missing #2

Closed torkencz closed 2 years ago

torkencz commented 2 years ago

Hi,

Very cool paper and thank you for making the code available. I've run into a problem where runFigR requires the package BuenRTools, which I cannot find anywhere. I think that the code from that was moved to utils.R, in which case line 142 of FigR_functions has to be modified so it doesn't require it.

vkartha commented 2 years ago

Hey Kristof,

   Thanks so much for your interest, and trying out the code. Sorry for

having left that in, you're absolutely right that dependency functions from one of my personal / private packages was moved over into utils.R, and is no longer needed. I believe the only place where it was specified was in the export packages parameter under the %dopar% parallelization setup. I have removed it from there, so hoping that did the trick. Let me know if you run into any issues. Also, we are close to finalising this code as a stand-alone R package (with documented examples and toy data), so stay tuned if that is of interest (I will update this repo to point to that, once it's ready!)

Best

On Sun, Nov 14, 2021 at 10:57 PM Kristof A Torkenczy < @.***> wrote:

Hi,

Very cool paper and thank you for making the code available. I've run into a problem where runFigR requires the package BuenRTools, which I cannot find anywhere. I think that the code from that was moved to utils.R, in which case line 142 of FigR_functions has to be modified so it doesn't require it.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/buenrostrolab/stimATAC_analyses_code/issues/2, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACJLS4X3LSD3VJJ7EZ7JGG3UMCAL3ANCNFSM5IAWSNWA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

-- Vinay Kartha Ph.D. Postdoctoral Associate Buenrostro lab Department of Stem Cell and Regenerative Biology Harvard University

torkencz commented 2 years ago

Thank you for your answer! Yes, now it works. I figured it out what was needed. I had to call the package SummarizedExperiment within the foreach loops along with sourcing the utils.R.

I have a separate question regarding the cell matching. What would constitute a good pairing of RNA and ATAC? I've run optMatch on some data and found some long range pairing on our UMAP image (10) that sorta make sense. Is there a way to be more strict to pair locally or having some distant pairing is OK? I've also played around with using L2 normalized CCA vs using the PCA of the Seurat ATAC-RNA integrated data and the latter seems to do better. Do you have any opinions on that. Again thank you for your help.

vkartha commented 2 years ago

Hi @torkencz , for future reference, please post separate issues / questions unrelated to the original issue on a separate page to make it more visible should others have similar questions!

Regarding the distant pairing - we have seen a very small % of distant pairing especially since we use the CCA/PCA space to do the kNN graph based pairing, and this doesn't have to equate in UMAP 2D space necessarily (but overall, it might), depending on how each was derived. If you want to be extra strict, you can always compute the euclidean distance for the same input ATAC/RNA matrices, between resulting pairs, and filter based on that (our experience again is that this isn't always the best determinant for picking "close" cells, and depends on your definition of "close".)

Regarding CCA vs PCA, thanks for letting us know that was the case for you, we did not extensively try out different dimensionality reduction methods as input to the pairing, and mainly focused on what was originally done in Seurat's CCA framework. I imagine the PCA is a lot faster, as one pro, and if you visualize the same in UMAP space, you can get a feel for how the two assays (ATAC/RNA) agree / co-embed, compared to the CCA. If one looks a lot better for any reason, then the pairing will also reflect that.

torkencz commented 2 years ago

Thank you for your answer. I am looking forward to the package!