Problem & Motivation
Dear friends,
scverse has spent a lot of effort into making zarr first class supported in AnnData. It's also the basis of all remote/cloud capabilities as well as https://github.com/scverse/annbatch. Currently, SingleCellCollection is primarily designed around h5ad (https://github.com/NVIDIA/bionemo-framework/blob/main/sub-packages/bionemo-scdl/src/bionemo/scdl/io/single_cell_collection.py#L87). There's a chance that future CELLxGENE versions will be shared in zarr format.
BioNeMo Framework Version
All
Category
API/Interface
Proposed Solution
I suggest that we at least mirror the h5ad API for zarr just like how AnnData mirrors its IO API.
Expected Benefits
Native zarr support and closer alignment with the ecosystem
Code Example
Problem & Motivation
Dear friends,
scverse has spent a lot of effort into making zarr first class supported in AnnData. It's also the basis of all remote/cloud capabilities as well as https://github.com/scverse/annbatch. Currently,
SingleCellCollectionis primarily designed around h5ad (https://github.com/NVIDIA/bionemo-framework/blob/main/sub-packages/bionemo-scdl/src/bionemo/scdl/io/single_cell_collection.py#L87). There's a chance that future CELLxGENE versions will be shared in zarr format.BioNeMo Framework Version
All
Category
API/Interface
Proposed Solution
I suggest that we at least mirror the h5ad API for zarr just like how AnnData mirrors its IO API.
Expected Benefits
Native zarr support and closer alignment with the ecosystem
Code Example