[Feature Request] Support for scVelo structured h5ad files for velocity-conos

[ACastanza]

Hello,

I noticed that there was a new(ish) tutorial for performing conos with velocity information. It would be great if this was capable of working with "generic" h5ad files with spliced and unspliced layers (just as you would use in the scVelo tutorial) by creating a panel of h5ad objects that had been read in with scVelo's scv.read() function.

I think allowing this to work with h5ad files would make it much more useful than requiring the specific dropEST pipeline process as there are multiple methods for extracting spliced and unspliced counts, and assembling them into the layers expected from the h5ad format is fairly trivial. This would also allow conos velocity to function on, for example, the combine lab's Alevin quantifications following their tutorial for producing scVelo compatible files.

Thanks!

[evanbiederstedt]

Hi @ACastanza

Thank you for using conos, and thank you for the feature request.

It would be great if this was capable of working with "generic" h5ad files with spliced and unspliced layers (just as you would use in the scVelo tutorial) by creating a panel of h5ad objects that had been read in with scVelo's scv.read() function.

This sounds like a pretty great feature.

The issue is, I don't think the lab has the bandwidth right now to implement this. It is certainly possibly nowadays to do this (given various AnnData interfaces), and others have written code for R to work with scVelo.

For the time being, the best approach is to label this with "help wanted" and ask the community to chip in to implement this. You are welcome to submit a PR yourself, of course!

Best, Evan