Exploring the Census Datasets table๏ƒ

This tutorial demonstrates basic use of the census_datasets dataframe that contains metadata of the Census source datasets. This metadata can be joined to the cell metadata dataframe (obs) via the column dataset_id,

Contents

  1. Fetching the datasets table.

  2. Fetching the expression data from a single dataset.

  3. Downloading the original source H5AD file of a dataset.

โš ๏ธ Note that the Census RNA data includes duplicate cells present across multiple datasets. Duplicate cells can be filtered in or out using the cell metadata variable is_primary_data which is described in the Census schema.

Fetching the datasets table๏ƒ

Each Census contains a top-level dataframe itemizing the datasets contained therein. You can read this into a pandas.DataFrame.

[1]:
import cellxgene_census

census = cellxgene_census.open_soma()
census_datasets = census["census_info"]["datasets"].read().concat().to_pandas()

# for convenience, indexing on the soma_joinid which links this to other census data.
census_datasets = census_datasets.set_index("soma_joinid")

census_datasets
The "stable" release is currently 2023-07-25. Specify 'census_version="2023-07-25"' in future calls to open_soma() to ensure data consistency.
[1]:
collection_id collection_name collection_doi dataset_id dataset_title dataset_h5ad_path dataset_total_cell_count
soma_joinid
0 e2c257e7-6f79-487c-b81c-39451cd4ab3c Spatial multiomics map of trophoblast developm... 10.1038/s41586-023-05869-0 f171db61-e57e-4535-a06a-35d8b6ef8f2b donor_p13_trophoblasts f171db61-e57e-4535-a06a-35d8b6ef8f2b.h5ad 31497
1 e2c257e7-6f79-487c-b81c-39451cd4ab3c Spatial multiomics map of trophoblast developm... 10.1038/s41586-023-05869-0 ecf2e08e-2032-4a9e-b466-b65b395f4a02 All donors trophoblasts ecf2e08e-2032-4a9e-b466-b65b395f4a02.h5ad 67070
2 e2c257e7-6f79-487c-b81c-39451cd4ab3c Spatial multiomics map of trophoblast developm... 10.1038/s41586-023-05869-0 74cff64f-9da9-4b2a-9b3b-8a04a1598040 All donors all cell states (in vivo) 74cff64f-9da9-4b2a-9b3b-8a04a1598040.h5ad 286326
3 f7cecffa-00b4-4560-a29a-8ad626b8ee08 Mapping single-cell transcriptomes in the intr... 10.1016/j.ccell.2022.11.001 5af90777-6760-4003-9dba-8f945fec6fdf Single-cell transcriptomic datasets of Renal c... 5af90777-6760-4003-9dba-8f945fec6fdf.h5ad 270855
4 3f50314f-bdc9-40c6-8e4a-b0901ebfbe4c Single-cell sequencing links multiregional imm... 10.1016/j.ccell.2021.03.007 bd65a70f-b274-4133-b9dd-0d1431b6af34 Single-cell sequencing links multiregional imm... bd65a70f-b274-4133-b9dd-0d1431b6af34.h5ad 167283
... ... ... ... ... ... ... ...
588 180bff9c-c8a5-4539-b13b-ddbc00d643e6 Molecular characterization of selectively vuln... 10.1038/s41593-020-00764-7 f9ad5649-f372-43e1-a3a8-423383e5a8a2 Molecular characterization of selectively vuln... f9ad5649-f372-43e1-a3a8-423383e5a8a2.h5ad 8168
589 a72afd53-ab92-4511-88da-252fb0e26b9a Single-cell atlas of peripheral immune respons... 10.1038/s41591-020-0944-y 456e8b9b-f872-488b-871d-94534090a865 Single-cell atlas of peripheral immune respons... 456e8b9b-f872-488b-871d-94534090a865.h5ad 44721
590 38833785-fac5-48fd-944a-0f62a4c23ed1 Construction of a human cell landscape at sing... 10.1038/s41586-020-2157-4 2adb1f8a-a6b1-4909-8ee8-484814e2d4bf Construction of a human cell landscape at sing... 2adb1f8a-a6b1-4909-8ee8-484814e2d4bf.h5ad 598266
591 5d445965-6f1a-4b68-ba3a-b8f765155d3a A molecular cell atlas of the human lung from ... 10.1038/s41586-020-2922-4 e04daea4-4412-45b5-989e-76a9be070a89 Krasnow Lab Human Lung Cell Atlas, Smart-seq2 e04daea4-4412-45b5-989e-76a9be070a89.h5ad 9409
592 5d445965-6f1a-4b68-ba3a-b8f765155d3a A molecular cell atlas of the human lung from ... 10.1038/s41586-020-2922-4 8c42cfd0-0b0a-46d5-910c-fc833d83c45e Krasnow Lab Human Lung Cell Atlas, 10X 8c42cfd0-0b0a-46d5-910c-fc833d83c45e.h5ad 65662

593 rows ร— 7 columns

The sum cells across all datasets should match the number of cells across all SOMA experiments (human, mouse).

[2]:
# Count cells across all experiments
all_experiments = (
    (organism_name, organism_experiment) for organism_name, organism_experiment in census["census_data"].items()
)
experiments_total_cells = 0
print("Count by experiment:")
for organism_name, organism_experiment in all_experiments:
    num_cells = len(organism_experiment.obs.read(column_names=["soma_joinid"]).concat().to_pandas())
    print(f"\t{num_cells} cells in {organism_name}")
    experiments_total_cells += num_cells

print(f"\nFound {experiments_total_cells} cells in all experiments.")

# Count cells across all datasets
print(f"Found {census_datasets.dataset_total_cell_count.sum()} cells in all datasets.")
Count by experiment:
        5255245 cells in mus_musculus
        56400873 cells in homo_sapiens

Found 61656118 cells in all experiments.
Found 61656118 cells in all datasets.

Fetching the expression data from a single dataset๏ƒ

Lets pick one dataset to slice out of the census, and turn into an AnnData in-memory object. This can be used with the ScanPy toolchain. You can also save this AnnData locally using the AnnData write API.

[3]:
census_datasets[census_datasets.dataset_id == "0bd1a1de-3aee-40e0-b2ec-86c7a30c7149"]
[3]:
collection_id collection_name collection_doi dataset_id dataset_title dataset_h5ad_path dataset_total_cell_count
soma_joinid
522 0b9d8a04-bb9d-44da-aa27-705bb65b54eb Tabula Muris Senis 10.1038/s41586-020-2496-1 0bd1a1de-3aee-40e0-b2ec-86c7a30c7149 Bone marrow - A single-cell transcriptomic atl... 0bd1a1de-3aee-40e0-b2ec-86c7a30c7149.h5ad 40220

Create a query on the mouse experiment, โ€œRNAโ€ measurement, for the dataset_id.

[4]:
adata = cellxgene_census.get_anndata(
    census, organism="Mus musculus", obs_value_filter="dataset_id == '0bd1a1de-3aee-40e0-b2ec-86c7a30c7149'"
)

adata
[4]:
AnnData object with n_obs ร— n_vars = 40220 ร— 52392
    obs: 'soma_joinid', 'dataset_id', 'assay', 'assay_ontology_term_id', 'cell_type', 'cell_type_ontology_term_id', 'development_stage', 'development_stage_ontology_term_id', 'disease', 'disease_ontology_term_id', 'donor_id', 'is_primary_data', 'self_reported_ethnicity', 'self_reported_ethnicity_ontology_term_id', 'sex', 'sex_ontology_term_id', 'suspension_type', 'tissue', 'tissue_ontology_term_id', 'tissue_general', 'tissue_general_ontology_term_id'
    var: 'soma_joinid', 'feature_id', 'feature_name', 'feature_length'

Downloading the original source H5AD file of a dataset.๏ƒ

You can download the original H5AD file for any given dataset. This is the same H5AD you can download from the CZ CELLxGENE Discover, and may contain additional data-submitter provided information which was not included in the Census.

To do this you can fetch the location in the cloud or directly download to your system using the cellxgene-census

[5]:
# Option 1: Direct download
cellxgene_census.download_source_h5ad(
    "0bd1a1de-3aee-40e0-b2ec-86c7a30c7149", to_path="Tabula_Muris_Senis-bone_marrow.h5ad"
)
[6]:
# Option 2: Get location and download via preferred method
uri = cellxgene_census.get_source_h5ad_uri("0bd1a1de-3aee-40e0-b2ec-86c7a30c7149")
uri

# you can now download the H5AD in shell via AWS CLI e.g. `aws s3 cp uri ./`
[6]:
{'uri': 's3://cellxgene-data-public/cell-census/2023-07-25/h5ads/0bd1a1de-3aee-40e0-b2ec-86c7a30c7149.h5ad',
 's3_region': 'us-west-2'}

Close the census

[7]:
census.close()