Closed vjcitn closed 9 months ago
Why do these functions need to live here? You could just as easily host them outside of the package, and I don't see how they would be useful to anyone outside of the Anvil use case.
Sure, I thought it might be helpful to have a hyperlinked README for the repo, and that the changes to ref.bib were advantageous for helping users get to the associated papers. The functions are there to help users; the reference to AnVIL was just an illustration of additional exposure.
https://github.com/vjcitn/scRNAseq#scrnaseq-package is the effect on repo README
A fleshed-out README is good but I don't think I want to see a hard-coded listing of datasets there, too much of a headache to keep in sync with the assets on ExperimentHub.
I think you might appreciate #42, which switches to the gypsum backend and adds titles, abstracts, PMIDs and other goodies to all datasets. This is available as a SQLite file (see the artifacts here) which can be more readily examined by users without adding all this extra machinery to parse the .bib
file.
Agreed that getting key metadata out of bibtex is not ideal. What do we have to do to get #42 merged? I will go back to our artifactdb dialogue but any specific steps welcome. Roughly, it sounds like enabling an AnnotationHub-like behavior for client packages in bioconductor, but using artifactdb.
If we can deal with Bioconductor/Contributions#3280, I will handle the rest of the ETL. The specific plan is to switch all of scRNAseq from ExperimentHub to use gypsum by default in the next release, and then we can have a look at its performance/cost in the wild. Especially if I open it up to third-party contributors who just supply SCEs.
It is done. Pushed to BioC-devel, let's see how it goes.
These enhancements are en route to the establishment of an AnVIL workspace devoted to the package. See https://app.terra.bio/#workspaces/landmarkanvil2/Bioconductor-Package-scRNAseq and scroll down for the hyperlinked table