def wu2020() -> AnnData:
"""\
Return the dataset from :cite:`Wu2020` as AnnData object.
200k cells, of which 100k have TCRs.
This is how the dataset was processed:
.. code-block:: python
{processing_code}
"""
url = "https://github.com/icbi-lab/scirpy/releases/download/d0.1.0/wu2020.h5ad"
filename = settings.datasetdir / "wu2020.h5ad"
adata = read(filename, backup_url=url)
return adata
def wu2020_3k() -> AnnData:
"""\
Return the dataset from :cite:`Wu2020` as AnnData object, downsampled
to 3000 TCR-containing cells.
This is how the dataset was processed:
.. code-block:: python
{processing_code}
"""
url = "https://github.com/icbi-lab/scirpy/releases/download/d0.1.0/wu2020_3k.h5ad"
filename = settings.datasetdir / "wu2020_3k.h5ad"
adata = read(filename, backup_url=url)
upgrade_schema(adata)
return adata
def wu2020_3k() -> AnnData:
"""\
Return the dataset from :cite:`Wu2020` as AnnData object, downsampled
to 3000 TCR-containing cells.
This is how the dataset was processed:
.. code-block:: python
{processing_code}
"""
# os.makedirs(settings.datasetdir, exist_ok=True)
# TODO host it on github or similar
url = "https://github.com/icbi-lab/scirpy/releases/download/d0.1.0/wu2020_3k.h5ad"
filename = settings.datasetdir / "wu2020_3k.h5ad"
adata = read(filename, backup_url=url)
return adata
def wu2020() -> AnnData:
"""
Return the dataset from [Wu2020]_ as AnnData object.
200k cells, of which 100k have TCRs.
This is how the dataset was processed:
```python
{processing_code}
```
"""
# os.makedirs(settings.datasetdir, exist_ok=True)
# TODO host it on github or similar
url = "https://github.com/icbi-lab/scirpy/releases/download/v0.1/wu2020.h5ad"
filename = settings.datasetdir / "wu2020.h5ad"
with _monkey_patch_tqdm():
adata = read(filename, backup_url=url)
return adata
def test_read_write_hdf5_sparse():
from scipy.sparse import csr_matrix
adata = AnnData(data=csr_matrix([[1, 0], [3, 0], [5, 6]]),
smp={
'row_names': ['name1', 'name2', 'name3'],
'sanno1': ['cat1', 'cat2', 'cat2'],
'sanno2': [2.1, 2.2, 2.3]
},
var={'vanno1': [3.1, 3.2]},
uns={
'sanno1_colors': ['#000000', '#FFFFFF'],
'uns2_sparse': csr_matrix([[1, 0], [3, 0]])
})
assert pd.api.types.is_string_dtype(adata.smp['sanno1'])
write('./test.h5', adata)
adata = read('./test.h5')
assert pd.api.types.is_categorical(adata.smp['sanno1'])
assert adata.smp.index.tolist() == ['name1', 'name2', 'name3']
assert adata.smp['sanno1'].cat.categories.tolist() == ['cat1', 'cat2']
def maynard2020() -> AnnData:
"""\
Return the dataset from :cite:`Maynard2020` as AnnData object.
21k cells from NSCLC profiled with Smart-seq2, of which 3,500 have :term:`TCRs<TCR>`
and 1,500 have :term:`BCRs<BCR>`.
The raw FASTQ files have been obtained from `PRJNA591860 <https://www.ebi.ac.uk/ena/browser/view/PRJNA591860>`__
and processed using the nf-core `Smart-seq2 pipeline <https://github.com/nf-core/smartseq2/>`__.
The processed files have been imported and transformed into an :class:`anndata.AnnData`
object using the following script:
.. code-block:: python
{processing_code}
"""
url = "https://github.com/icbi-lab/scirpy/releases/download/d0.1.0/maynard2020.h5ad"
filename = settings.datasetdir / "maynard2020.h5ad"
adata = read(filename, backup_url=url)
return adata
def test_read_write_hdf5():
adata = AnnData(
data=np.array([[1, 0], [3, 0], [5, 6]]),
smp={
'row_names': ['name1', 'name2', 'name3'],
'sanno1': ['cat1', 'cat2', 'cat2'], # categorical anno
'sanno2': ['s1', 's2', 's3'], # string annotation
'sanno3': [2.1, 2.2, 2.3]
}, # float annotation
var={'vanno1': [3.1, 3.2]},
uns={
'sanno1_colors': ['#000000', '#FFFFFF'],
'uns2': ['some annotation']
})
assert pd.api.types.is_string_dtype(adata.smp['sanno1'])
write('./test.h5', adata)
adata = read('./test.h5')
assert pd.api.types.is_categorical(adata.smp['sanno1'])
assert pd.api.types.is_string_dtype(adata.smp['sanno2'])
assert adata.smp.index.tolist() == ['name1', 'name2', 'name3']
assert adata.smp['sanno1'].cat.categories.tolist() == ['cat1', 'cat2']
def maynard2020_3k() -> AnnData:
"""\
Return the dataset from :cite:`Maynard2020` as AnnData object, downsampled
to 3000 cells.
In brief, this data set was processed as follows:
* raw data downloaded from ENA
* gene expression quantified using Salmon and the nf-core/rnaseq pipeline.
* basic quality control (min_counts=20k, max_counts=5M, min_genes=1k, max_mitochondrial_fraction=0.2)
* filtered to 6000 HVG using `sc.pp.highly_variable_genes(..., flavor="seurat_v3")`
* raw counts processed using scVI, providing sample information as batch key.
* cell types manually annotated based on marker genes and leiden clustering and subclustering.
* downsampled to 3000 cells.
`adata.X` contains the `log1p` transformed, cpm-normalized raw counts.
The `scVI` latent representation is stored in `adata.obsm["X_scVI"]`.
A UMAP for the 3000 cells is precomputed.
"""
url = "https://github.com/icbi-lab/infercnvpy/releases/download/d0.1.0/maynard2020_3k.h5ad"
filename = settings.datasetdir / "maynard2020_3k.h5ad"
adata = read(filename, backup_url=url)
return adata