Add CLI argument to launch for switching between .X and .raw

[sidneymbell]

Thanks so much to @gokceneraslan (#676) and @Marius1311 (#542) for flagging the related issues of how to visualize .raw versus .X (and making it clear which one you're looking at!).

Making a new issue to offer a proposed solution for discussion.

It's rather confusing and/or misleading to visualize normalized and unnormalized values in the same view. One solution would be to toggle back and forth between viewing .raw and .X within the app (as suggested by @gokceneraslan :).

However, the frontend currently only accommodates a single matrix of values. Under the hood, toggling would be effectively the same as relaunching.

I'd propose to just add a command line argument to launch that allows the user to specify whether they want to load the raw count values from .raw or the normalized values from .X (the default).

Would this adequately solve the needs presented in #676 and #542? All feedback and thoughts welcome!

[Marius1311]

Dear @sidneymbell , that sounds good to me!

[gokceneraslan]

Sounds good to me, too. Ideally, the gene search should always be performed in .raw since gene subsetting in Scanpy (i.e. ad = ad[:, somegenes] or sc.pp.filter_genes or sc.pp.highly_variable_genes(subset=True)) filters out genes in X but always keep all genes in raw.X. But if it's challenging technically, specifying from the command line is also ok.

What about the layers? I think it also makes sense to select a layer from the command line. What do you think?

[alokito]

Anecdotally some users say being able to toggle between raw and normalized values would be ideal, and that Seurat recommends performing differential expression analysis on the raw values.

[signechambers1]

Fixed via #1030