Better support R users

tl;dr – I’d love thoughts and discussion around this. I suspect this has several components to it; one big component of this is input data formats. I’ve taken some notes below about what (I think?) that landscape looks like based on some initial investigations.

In targeting the loop between experimental and computational users, it would allow us to support many more scientists if launch were able to support the output of seurat and bioconductor.

For improving data ingress, I’d vote for implementing this in launch rather than prepare simply because if someone is already working in seurat/bioconductor (i.e., in an analytical loop), I assume they don’t need to use prepare to do normalization, embedding, etc. I think this is also where we’d most likely be intercepting pairs of experimental/computational users.

This does get a little tricky. In exploring and consulting with Angela Pisco and Olga Botvinnik, it appears that both seurat and bioconductor mostly just output rds and rdata objects. While there are converters between the seurat <–> SingleCellExperiment R objects --> h5ad, these are still under development and nontrivial to work with.

In the future, I suspect supporting launch from rdata/rds will be necessary. But, it does seem like a larger and more complex chunk of work than I had anticipated due to the absence of an intermediate data file structure.

In the meantime, I’d suggest implementing support for other file formats already supported by anndata.read as detailed here.

[csweaver]

I think that supporting the R ecosystem and supporting input from other file formats are going to be two separate tickets. For this ticket the title matchings the supporting input, but the body matches supporting R. And it seems like supporting R is a more involved process which deserves its own prioritization for roadmapping and releases.

[signechambers1]

Closing as a duplicate of #1841