cellink.io.read_pgen_zarr

cellink.io.read_pgen_zarr(store)

Lazily read an AnnData Zarr v3 store written by stream_pgen_to_zarr.

This function reconstructs an anndata.AnnData object from a Zarr store while keeping the primary data matrix (X) backed by Dask arrays. It is designed for large genotype matrices that cannot be loaded fully into memory.

The reader preserves:

• Dense X stored as a Zarr array (returned as a Dask-backed array)

• Sparse matrices (CSR/CSC)

• DataFrames (obs, var)

• Awkward arrays

• Standard AnnData container structure

Parameters:

store (str or pathlib.Path) – Path to a Zarr directory created by stream_pgen_to_zarr or a compatible AnnData Zarr v3 store.

Return type:

AnnData

Returns:

anndata.AnnData AnnData object with: - X as a Dask-backed array (for dense storage) - obs and var as pandas DataFrames - empty container groups (uns, obsm, varm, layers, etc.)

if present in the store

Notes

• The returned object is lazy when X is dense. Computation is triggered only when .compute() or in-memory materialization is requested.

• For sparse X written via stream_pgen_to_zarr(..., sparse=True), the matrix is loaded as a SciPy sparse matrix.

• This function relies on AnnData’s experimental dispatched I/O API.

Examples

>>> import cellink
>>> adata = cellink.io.read_pgen_zarr("genotypes.zarr")
>>> adata
AnnData object with n_obs x n_vars = ...

>>> # Trigger computation
>>> X = adata.X.compute()