Curators and Data Consumers want Discover API to return AnnData elements from a dataset

[brianraymor]

Context

@jahilton has asked for obs and var elements to be returned.

See scenario from @emdann :

I often use census to analyse tables of cell- and sample-level metadata before downloading h5ads, to decide whether I want to include or exclude a dataset or subset of cells from analysis. Often info in the author metadata informs that decision, but right now I can only access it by downloading the whole h5ad for a dataset. It would be great to be able to browse the column names of the author metadata for a certain dataset_id, or even better to download the full adata.obs for a dataset_id without downloading all the count matrices.

Also see @jahilton comment.

---

This code fragment using anndata.experimental which has been previously shared:

import h5py
import fsspec
from anndata.experimental import read_elem
# or from anndata.io import read_elem for 0.11

fs = fsspec.filesystem('s3', anon=True)

with h5py.File(
    fs.open("cellxgene-census-public-us-west-2/cell-census/2024-07-01/h5ads/0041b9c3-6a49-4bf7-8514-9bc7190067a7.h5ad")
) as f:
    obs = read_elem(f["obs"])