Update get-started with resolutions of questions from demo

[falexwolf]

This is meant to resolve the two below issues:

• In get-started, use .var to annotate genes #103
• Add an example for slicing census / tiledbsoma #108

[falexwolf]

Why does this fail? I just copied from Robrecht's notebook:

https://lamin.ai/laminlabs/lamindata/transform/y0MIazBuljv70002

[falexwolf]

I also added an exemplary of Census to address the other important question that came up during the demo.

There are ways to make this more natively supported, but this isn't urgend. I made an issue for this:

• Enable artifact.open() for tiledbsoma backends #114

[falexwolf]

Can one of you please finalize this, @lazappi or @rcannood? 🙇

[lazappi]

I also added an exemplary of Census to address the other important question that came up during the demo.

Does this fit in this vignette? It doesn't seem like part of the introductory workflow and doesn't use {laminr}. Could we just point someone to the documentation for the {cellxgene.census} package if they want to know how to do this?

[falexwolf]

Does this fit in this vignette? It doesn't seem like part of the introductory workflow and doesn't use {laminr}. Could we just point someone to the documentation for the {cellxgene.census} package if they want to know how to do this?

I agree it doesn't generally fit into this vignette, but right now the main audience for this is the person who asked the question; so it should be featured here on this page.

Once the below issue is addressed it will fit well, though:

• Enable artifact.open() for tiledbsoma backends #114

So: Let's please keep it here for now but not for long in the present form. The section will be replaced with the $open() call in the upcoming weeks, which is fundamentally different from $load() and $cache() and hence merits to be shown in get-started.

[falexwolf]

Can you please help to get this to pass, @lazappi? It's going to take you much less time than I. I'm simply not familiar enough.

[lazappi]

I adjusted the chunk a bit but there are some bigger issues:

1. The {cellxgene.census} package is currently not working (see [R] cellxgene-census PermanentRedirect S3 error chanzuckerberg/cellxgene-census#1261). I haven't been able to run the code manually even with the workarounds listed here. (The Python package is having similar issues DoesNotExistError: 's3://cellxgene-census-public-us-west-2/cell-census/2023-12-15/soma/' does not exist chanzuckerberg/cellxgene-census#1313)
2. {cellxgene.census} is not available from CRAN/Bioconductor and only works on Unix. This means it is difficult for people to install and we can't really add it as a dependency so the CI will fail without some workarounds and CRAN probably won't allow it.

I think the best we could do at the moment is to have the code there but not evaluated, I'm not sure how useful that is if people can't install/use the package anyway though. There is a Bioconductor {cellxgenedp} package that could maybe be used instead but I'm not sure if it does what you are looking for.

[falexwolf]

Thanks, Luke!

I think the best we could do at the moment is to have the code there but not evaluated,

I agree.

I'm not sure how useful that is if people can't install/use the package anyway though

It's useful to illustrate the point of streaming a big array to get a Seurat object vs. caching a small array locally. This is important to convey to the current target audience, even if the caveats that you're listing make clear that this is currently problematic. I'm expecting these issues will get resolved by CZI though.

If it's installable on UNIX that's probably good enough. My understanding is they all work on Linux servers in the cloud.

[rcannood]

LGTM