Querying data using the gget cellxgene module

By Laura Luebbert, lauraluebbert@caltech.edu.

gget is a free, open-source command-line tool and Python package that enables efficient querying of genomic databases. gget consists of a collection of separate but interoperable modules, each designed to facilitate one type of database querying in a single line of code.

The gget cellxgene module builds on the CZ CELLxGENE Discover Census to query data from CZ CELLxGENE Discover. This notebook briefly introduces the gget cellxgene module by providing one simple example for each supported query type.

If you use gget cellxgene in a publication, please cite gget in addition to citing CZ CELLxGENE.

You can also open this notebook in Google Colab.

Contents

1. Install gget.

2. Fetch an AnnData object by selecting gene(s), tissue(s) and cell type(s).

3. Plot a dot plot similar to those shown on the CZ CELLxGENE Discover Gene Expression.

4. Fetch only cell metadata (corresponds to AnnData.obs).

5. Use gget cellxgene from the command line.

Install gget and set up cellxgene module

# The cellxgene module was added to gget in version 0.25.7
!pip install -q gget >=0.25.7

import gget

gget.setup("cellxgene")

[notice] A new release of pip available: 22.3.1 -> 23.2.1
[notice] To update, run: pip install --upgrade pip

Fri Jul 28 16:16:17 2023 INFO Installing cellxgene-census package (requires pip).
Fri Jul 28 16:16:22 2023 INFO cellxgene_census installed succesfully.

# Display all options of the cellxgene gget module
help(gget.cellxgene)

Help on function cellxgene in module gget.gget_cellxgene:

cellxgene(species='homo_sapiens', gene=None, ensembl=False, column_names=['dataset_id', 'assay', 'suspension_type', 'sex', 'tissue_general', 'tissue', 'cell_type'], meta_only=False, tissue=None, cell_type=None, development_stage=None, disease=None, sex=None, is_primary_data=True, dataset_id=None, tissue_general_ontology_term_id=None, tissue_general=None, assay_ontology_term_id=None, assay=None, cell_type_ontology_term_id=None, development_stage_ontology_term_id=None, disease_ontology_term_id=None, donor_id=None, self_reported_ethnicity_ontology_term_id=None, self_reported_ethnicity=None, sex_ontology_term_id=None, suspension_type=None, tissue_ontology_term_id=None, census_version='stable', verbose=True, out=None)
    Query data from CZ CELLxGENE Discover (https://cellxgene.cziscience.com/) using the
    CZ CELLxGENE Discover Census (https://github.com/chanzuckerberg/cellxgene-census).

    NOTE: Querying large datasets requires a large amount of RAM. Use the cell metadata attributes
    to define the (sub)dataset of interest.
    The CZ CELLxGENE Discover Census recommends >16 GB of memory and a >5 Mbps internet connection.

    General args:
        - species        Choice of 'homo_sapiens' or 'mus_musculus'. Default: 'homo_sapiens'.
        - gene           Str or list of gene name(s) or Ensembl ID(s), e.g. ['ACE2', 'SLC5A1'] or ['ENSG00000130234', 'ENSG00000100170']. Default: None.
                         NOTE: Set ensembl=True when providing Ensembl ID(s) instead of gene name(s).
                         See https://cellxgene.cziscience.com/gene-expression for examples of available genes.
        - ensembl        True/False (default: False). Set to True when genes are provided as Ensembl IDs.
        - column_names   List of metadata columns to return (stored in AnnData.obs when meta_only=False).
                         Default: ["dataset_id", "assay", "suspension_type", "sex", "tissue_general", "tissue", "cell_type"]
                         For more options see: https://api.cellxgene.cziscience.com/curation/ui/#/ -> Schemas -> dataset
        - meta_only      True/False (default: False). If True, returns only metadata dataframe (corresponds to AnnData.obs).
        - census_version Str defining version of Census, e.g. "2023-05-15" or "latest" or "stable". Default: "stable".
        - verbose        True/False whether to print progress information. Default True.
        - out            If provided, saves the generated AnnData h5ad (or csv when meta_only=True) file with the specified path. Default: None.

    Cell metadata attributes:
        - tissue                          Str or list of tissue(s), e.g. ['lung', 'blood']. Default: None.
                                          See https://cellxgene.cziscience.com/gene-expression for examples of available tissues.
        - cell_type                       Str or list of celltype(s), e.g. ['mucus secreting cell', 'neuroendocrine cell']. Default: None.
                                          See https://cellxgene.cziscience.com/gene-expression and select a tissue to see examples of available celltypes.
        - development_stage               Str or list of development stage(s). Default: None.
        - disease                         Str or list of disease(s). Default: None.
        - sex                             Str or list of sex(es), e.g. 'female'. Default: None.
        - is_primary_data                 True/False (default: True). If True, returns only the canonical instance of the cellular observation.
                                          This is commonly set to False for meta-analyses reusing data or for secondary views of data.
        - dataset_id                      Str or list of CELLxGENE dataset ID(s). Default: None.
        - tissue_general_ontology_term_id Str or list of high-level tissue UBERON ID(s). Default: None.
                                          Also see: https://github.com/chanzuckerberg/single-cell-data-portal/blob/9b94ccb0a2e0a8f6182b213aa4852c491f6f6aff/backend/wmg/data/tissue_mapper.py
        - tissue_general                  Str or list of high-level tissue label(s). Default: None.
                                          Also see: https://github.com/chanzuckerberg/single-cell-data-portal/blob/9b94ccb0a2e0a8f6182b213aa4852c491f6f6aff/backend/wmg/data/tissue_mapper.py
        - tissue_ontology_term_id         Str or list of tissue ontology term ID(s) as defined in the CELLxGENE dataset schema. Default: None.
        - assay_ontology_term_id          Str or list of assay ontology term ID(s) as defined in the CELLxGENE dataset schema. Default: None.
        - assay                           Str or list of assay(s) as defined in the CELLxGENE dataset schema. Default: None.
        - cell_type_ontology_term_id      Str or list of celltype ontology term ID(s) as defined in the CELLxGENE dataset schema. Default: None.
        - development_stage_ontology_term_id        Str or list of development stage ontology term ID(s) as defined in the CELLxGENE dataset schema. Default: None.
        - disease_ontology_term_id        Str or list of disease ontology term ID(s) as defined in the CELLxGENE dataset schema. Default: None.
        - donor_id                        Str or list of donor ID(s) as defined in the CELLxGENE dataset schema. Default: None.
        - self_reported_ethnicity_ontology_term_id  Str or list of self reported ethnicity ontology ID(s) as defined in the CELLxGENE dataset schema. Default: None.
        - self_reported_ethnicity         Str or list of self reported ethnicity as defined in the CELLxGENE dataset schema. Default: None.
        - sex_ontology_term_id            Str or list of sex ontology ID(s) as defined in the CELLxGENE dataset schema. Default: None.
        - suspension_type                 Str or list of suspension type(s) as defined in the CELLxGENE dataset schema. Default: None.

    Returns AnnData object (when meta_only=False) or dataframe (when meta_only=True).

Fetch an AnnData object by selecting gene(s), tissue(s) and cell type(s)

You can use all of the options listed above to filter for data of interest. Here, we will demonstrate the module by fetching a small dataset containing only three genes and two lung cell types:

# Fetch AnnData object based on specified genes, tissue and cell types
adata = gget.cellxgene(
    gene=["ACE2", "ABCA1", "SLC5A1"], tissue="lung", cell_type=["mucus secreting cell", "neuroendocrine cell"]
)

Fri Jul 28 16:16:22 2023 INFO Fetching AnnData object from CZ CELLxGENE Discover. This might take a few minutes...
The "stable" release is currently 2023-07-25. Specify 'census_version="2023-07-25"' in future calls to open_soma() to ensure data consistency.
Fri Jul 28 16:16:22 2023 INFO The "stable" release is currently 2023-07-25. Specify 'census_version="2023-07-25"' in future calls to open_soma() to ensure data consistency.

Let’s look at some of the features of the AnnData object we just fetched:

adata

AnnData object with n_obs × n_vars = 3679 × 3
    obs: 'dataset_id', 'assay', 'suspension_type', 'sex', 'tissue_general', 'tissue', 'cell_type', 'is_primary_data'
    var: 'soma_joinid', 'feature_id', 'feature_name', 'feature_length'

A few thousand cells from CZ CELLxGENE Discover matched the filters specified above and their ACE2, ABCA1, and SLC5A1 expression matrix in lung mucus secreting and neuroendocrine cells was fetched. The .var and .obs layers contain additional information about each gene and cell, respectively:

adata.var

	soma_joinid	feature_id	feature_name	feature_length
0	38	ENSG00000165029	ABCA1	11343
1	5332	ENSG00000130234	ACE2	9739
2	24539	ENSG00000100170	SLC5A1	5081

adata.obs

	dataset_id	assay	suspension_type	sex	tissue_general	tissue	cell_type	is_primary_data
0	9f222629-9e39-47d0-b83f-e08d610c7479	10x 3' v2	cell	unknown	lung	lung	mucus secreting cell	True
1	9f222629-9e39-47d0-b83f-e08d610c7479	10x 3' v2	cell	unknown	lung	lung	mucus secreting cell	True
2	9f222629-9e39-47d0-b83f-e08d610c7479	10x 3' v2	cell	unknown	lung	lung	mucus secreting cell	True
3	9f222629-9e39-47d0-b83f-e08d610c7479	10x 3' v2	cell	unknown	lung	lung	mucus secreting cell	True
4	9f222629-9e39-47d0-b83f-e08d610c7479	10x 3' v2	cell	unknown	lung	lung	mucus secreting cell	True
...	...	...	...	...	...	...	...	...
3674	8c42cfd0-0b0a-46d5-910c-fc833d83c45e	10x 3' v2	cell	female	lung	lung	mucus secreting cell	True
3675	8c42cfd0-0b0a-46d5-910c-fc833d83c45e	10x 3' v2	cell	female	lung	lung	mucus secreting cell	True
3676	8c42cfd0-0b0a-46d5-910c-fc833d83c45e	10x 3' v2	cell	female	lung	lung	mucus secreting cell	True
3677	8c42cfd0-0b0a-46d5-910c-fc833d83c45e	10x 3' v2	cell	female	lung	lung	mucus secreting cell	True
3678	8c42cfd0-0b0a-46d5-910c-fc833d83c45e	10x 3' v2	cell	female	lung	lung	mucus secreting cell	True

3679 rows × 8 columns

Plot a dot plot similar to those shown on the CZ CELLxGENE Discover Gene Expression

Using the data we just fetched, we can plot a dot plot using scanpy:

import scanpy as sc

# retina increases the resolution of plots displayed in notebooks
%config InlineBackend.figure_format="retina"

sc.pl.dotplot(adata, adata.var["feature_name"].values, groupby="cell_type", gene_symbols="feature_name")

... storing 'dataset_id' as categorical
... storing 'assay' as categorical
... storing 'suspension_type' as categorical
... storing 'sex' as categorical
... storing 'tissue_general' as categorical
... storing 'tissue' as categorical
... storing 'cell_type' as categorical

Fetch only cell metadata (corresponds to AnnData.obs)

By setting meta_only=True and again filtering by the cell metadata attributes listed above, you can also fetch only the cell metadata:

df = gget.cellxgene(
    meta_only=True,
    census_version="2023-05-15",  # Specify Census version for reproducibility over time
    gene="ENSMUSG00000015405",
    ensembl=True,  # Setting 'ensembl=True' here since the gene is passed as an Ensembl ID
    tissue="lung",
    species="mus_musculus",  # Let's switch up the species
)

df

Fri Jul 28 16:16:44 2023 INFO Fetching metadata from CZ CELLxGENE Discover...

	dataset_id	assay	suspension_type	sex	tissue_general	tissue	cell_type	is_primary_data
0	047d57f2-4d14-45de-aa98-336c6f583750	10x 3' v2	cell	unknown	lung	lung	mesenchymal stem cell	True
1	047d57f2-4d14-45de-aa98-336c6f583750	10x 3' v2	cell	unknown	lung	lung	progenitor cell	True
2	047d57f2-4d14-45de-aa98-336c6f583750	10x 3' v2	cell	unknown	lung	lung	mesenchymal cell	True
3	047d57f2-4d14-45de-aa98-336c6f583750	10x 3' v2	cell	unknown	lung	lung	mesenchymal stem cell	True
4	047d57f2-4d14-45de-aa98-336c6f583750	10x 3' v2	cell	unknown	lung	lung	mesenchymal cell	True
...	...	...	...	...	...	...	...	...
97547	48b37086-25f7-4ecd-be66-f5bb378e3aea	10x 3' v2	cell	male	lung	lung	fibroblast of lung	True
97548	48b37086-25f7-4ecd-be66-f5bb378e3aea	10x 3' v2	cell	male	lung	lung	natural killer cell	True
97549	48b37086-25f7-4ecd-be66-f5bb378e3aea	10x 3' v2	cell	male	lung	lung	pulmonary interstitial fibroblast	True
97550	48b37086-25f7-4ecd-be66-f5bb378e3aea	10x 3' v2	cell	male	lung	lung	adventitial cell	True
97551	48b37086-25f7-4ecd-be66-f5bb378e3aea	10x 3' v2	cell	male	lung	lung	fibroblast of lung	True

97552 rows × 8 columns

Use gget cellxgene from the command line

All gget modules support use from the command line. Note that the command line interface requires the -o/--out argument to specify a path to save the fetched data. Here are the command line versions of the queries demonstrated above:

# # Fetch AnnData object based on specified genes, tissue and cell types
# !gget cellxgene --gene ACE2 ABCA1 SLC5A1 --tissue lung --cell_type 'mucus secreting cell' 'neuroendocrine cell' -o example_adata.h5ad

# # Fetch only metadata
# !gget cellxgene --meta_only --gene ENSMUSG00000015405 --ensembl --tissue lung --species mus_musculus -o example_meta.csv