def centrality_scores(
adata: AnnData,
cluster_key: str,
score: Optional[Union[str, Iterable[str]]] = None,
connectivity_key: Optional[str] = None,
copy: bool = False,
n_jobs: Optional[int] = None,
backend: str = "loky",
show_progress_bar: bool = False,
) -> Optional[pd.DataFrame]:
"""
Compute centrality scores per cluster or cell type.
Inspired by usage in Gene Regulatory Networks (GRNs) in :cite:`celloracle`.
Parameters
----------
%(adata)s
%(cluster_key)s
score
Centrality measures as described in :class:`networkx.algorithms.centrality` :cite:`networkx`.
If `None`, use all the options below. Valid options are:
- `{c.CLOSENESS.s!r}` - measure of how close the group is to other nodes.
- `{c.CLUSTERING.s!r}` - measure of the degree to which nodes cluster together.
- `{c.DEGREE.s!r}` - fraction of non-group members connected to group members.
%(conn_key)s
%(copy)s
%(parallelize)s
Returns
-------
If ``copy = True``, returns a :class:`pandas.DataFrame`. Otherwise, modifies the ``adata`` with the following key:
- :attr:`anndata.AnnData.uns` ``['{{cluster_key}}_centrality_scores']`` - the centrality scores,
as mentioned above.
"""
connectivity_key = Key.obsp.spatial_conn(connectivity_key)
_assert_categorical_obs(adata, cluster_key)
_assert_connectivity_key(adata, connectivity_key)
if isinstance(score, (str, Centrality)):
centrality = [score]
elif score is None:
centrality = [c.s for c in Centrality]
centralities = [Centrality(c) for c in centrality]
graph = nx.from_scipy_sparse_matrix(adata.obsp[connectivity_key])
cat = adata.obs[cluster_key].cat.categories.values
clusters = adata.obs[cluster_key].values
fun_dict = {}
for c in centralities:
if c == Centrality.CLOSENESS:
fun_dict[c.s] = partial(
nx.algorithms.centrality.group_closeness_centrality, graph)
elif c == Centrality.DEGREE:
fun_dict[c.s] = partial(
nx.algorithms.centrality.group_degree_centrality, graph)
elif c == Centrality.CLUSTERING:
fun_dict[c.s] = partial(nx.algorithms.cluster.average_clustering,
graph)
else:
raise NotImplementedError(
f"Centrality `{c}` is not yet implemented.")
n_jobs = _get_n_cores(n_jobs)
start = logg.info(
f"Calculating centralities `{centralities}` using `{n_jobs}` core(s)")
res_list = []
for k, v in fun_dict.items():
df = parallelize(
_centrality_scores_helper,
collection=cat,
extractor=pd.concat,
n_jobs=n_jobs,
backend=backend,
show_progress_bar=show_progress_bar,
)(clusters=clusters, fun=v, method=k)
res_list.append(df)
df = pd.concat(res_list, axis=1)
if copy:
return df
_save_data(adata,
attr="uns",
key=Key.uns.centrality_scores(cluster_key),
data=df,
time=start)
def _analysis(
data: pd.DataFrame,
interactions: np.ndarray,
interaction_clusters: np.ndarray,
threshold: float = 0.1,
n_perms: int = 1000,
seed: Optional[int] = None,
n_jobs: int = 1,
numba_parallel: Optional[bool] = None,
**kwargs: Any,
) -> TempResult:
"""
Run the analysis as described in :cite:`cellphonedb`.
This function runs the mean, percent and shuffled analysis.
Parameters
----------
data
Array of shape `(n_cells, n_genes)`.
interactions
Array of shape `(n_interactions, 2)`.
interaction_clusters
Array of shape `(n_interaction_clusters, 2)`.
threshold
Percentage threshold for removing lowly expressed genes in clusters.
%(n_perms)s
%(seed)s
n_jobs
Number of parallel jobs to launch.
numba_parallel
Whether to use :class:`numba.prange` or not. If `None`, it's determined automatically.
kwargs
Keyword arguments for :func:`squidpy._utils.parallelize`, such as ``n_jobs`` or ``backend``.
Returns
-------
Tuple of the following format:
- `'means'` - array of shape `(n_interactions, n_interaction_clusters)` containing the means.
- `'pvalues'` - array of shape `(n_interactions, n_interaction_clusters)` containing the p-values.
"""
def extractor(res: Sequence[TempResult]) -> TempResult:
assert len(
res
) == n_jobs, f"Expected to find `{n_jobs}` results, found `{len(res)}`."
meanss: List[np.ndarray] = [
r.means for r in res if r.means is not None
]
assert len(
meanss
) == 1, f"Only `1` job should've calculated the means, but found `{len(meanss)}`."
means = meanss[0]
if TYPE_CHECKING:
assert isinstance(means, np.ndarray)
pvalues = np.sum([r.pvalues for r in res if r.pvalues is not None],
axis=0) / float(n_perms)
assert means.shape == pvalues.shape, f"Means and p-values differ in shape: `{means.shape}`, `{pvalues.shape}`."
return TempResult(means=means, pvalues=pvalues)
groups = data.groupby("clusters")
clustering = np.array(data["clusters"].values, dtype=np.int32)
mean = groups.mean().values.T # (n_genes, n_clusters)
mask = groups.apply(lambda c: (
(c > 0).sum() / len(c)) >= threshold).values.T # (n_genes, n_clusters)
# (n_cells, n_genes)
data = np.array(data[data.columns.difference(["clusters"])].values,
dtype=np.float64,
order="C")
# all 3 should be C contiguous
return parallelize( # type: ignore[no-any-return]
_analysis_helper,
np.arange(n_perms, dtype=np.int32),
n_jobs=n_jobs,
unit="permutation",
extractor=extractor,
**kwargs,
)(
data,
mean,
mask,
interactions,
interaction_clusters=interaction_clusters,
clustering=clustering,
seed=seed,
numba_parallel=numba_parallel,
)
def nhood_enrichment(
adata: AnnData,
cluster_key: str,
connectivity_key: Optional[str] = None,
n_perms: int = 1000,
numba_parallel: bool = False,
seed: Optional[int] = None,
copy: bool = False,
n_jobs: Optional[int] = None,
backend: str = "loky",
show_progress_bar: bool = True,
) -> Optional[Tuple[np.ndarray, np.ndarray]]:
"""
Compute neighborhood enrichment by permutation test.
Parameters
----------
%(adata)s
%(cluster_key)s
%(conn_key)s
%(n_perms)s
%(numba_parallel)s
%(seed)s
%(copy)s
%(parallelize)s
Returns
-------
If ``copy = True``, returns a :class:`tuple` with the z-score and the enrichment count.
Otherwise, modifies the ``adata`` with the following keys:
- :attr:`anndata.AnnData.uns` ``['{cluster_key}_nhood_enrichment']['zscore']`` - the enrichment z-score.
- :attr:`anndata.AnnData.uns` ``['{cluster_key}_nhood_enrichment']['count']`` - the enrichment count.
"""
connectivity_key = Key.obsp.spatial_conn(connectivity_key)
_assert_categorical_obs(adata, cluster_key)
_assert_connectivity_key(adata, connectivity_key)
_assert_positive(n_perms, name="n_perms")
adj = adata.obsp[connectivity_key]
original_clust = adata.obs[cluster_key]
clust_map = {
v: i
for i, v in enumerate(original_clust.cat.categories.values)
} # map categories
int_clust = np.array([clust_map[c] for c in original_clust], dtype=ndt)
indices, indptr = (adj.indices.astype(ndt), adj.indptr.astype(ndt))
n_cls = len(clust_map)
_test = _create_function(n_cls, parallel=numba_parallel)
count = _test(indices, indptr, int_clust)
n_jobs = _get_n_cores(n_jobs)
start = logg.info(
f"Calculating neighborhood enrichment using `{n_jobs}` core(s)")
perms = parallelize(
_nhood_enrichment_helper,
collection=np.arange(n_perms),
extractor=np.vstack,
n_jobs=n_jobs,
backend=backend,
show_progress_bar=show_progress_bar,
)(callback=_test,
indices=indices,
indptr=indptr,
int_clust=int_clust,
n_cls=n_cls,
seed=seed)
zscore = (count - perms.mean(axis=0)) / perms.std(axis=0)
if copy:
return zscore, count
_save_data(
adata,
attr="uns",
key=Key.uns.nhood_enrichment(cluster_key),
data={
"zscore": zscore,
"count": count
},
time=start,
)
def moran(
adata: AnnData,
connectivity_key: str = Key.obsp.spatial_conn(),
genes: Optional[Union[str, Sequence[str]]] = None,
transformation: Literal["r", "B", "D", "U", "V"] = "r",
n_perms: int = 1000,
corr_method: Optional[str] = "fdr_bh",
layer: Optional[str] = None,
seed: Optional[int] = None,
copy: bool = False,
n_jobs: Optional[int] = None,
backend: str = "loky",
show_progress_bar: bool = True,
) -> Optional[pd.DataFrame]:
"""
Calculate Moran’s I Global Autocorrelation Statistic.
Parameters
----------
%(adata)s
%(conn_key)s
genes
List of gene names, as stored in :attr:`anndata.AnnData.var_names`, used to compute Moran's I statistics
:cite:`pysal`.
If `None`, it's computed :attr:`anndata.AnnData.var` ``['highly_variable']``, if present. Otherwise,
it's computed for all genes.
transformation
Transformation to be used, as reported in :class:`esda.Moran`. Default is `"r"`, row-standardized.
%(n_perms)s
%(corr_method)s
layer
Layer in :attr:`anndata.AnnData.layers` to use. If `None`, use :attr:`anndata.AnnData.X`.
%(seed)s
%(copy)s
%(parallelize)s
Returns
-------
If ``copy = True``, returns a :class:`pandas.DataFrame` with the following keys:
- `'I'` - Moran's I statistic.
- `'pval_sim'` - p-value based on permutations.
- `'VI_sim'` - variance of `'I'` from permutations.
- `'pval_sim_{{corr_method}}'` - the corrected p-values if ``corr_method != None`` .
Otherwise, modifies the ``adata`` with the following key:
- :attr:`anndata.AnnData.uns` ``['moranI']`` - the above mentioned dataframe.
"""
if esda is None or libpysal is None:
raise ImportError(
"Please install `esda` and `libpysal` as `pip install esda libpysal`."
)
_assert_positive(n_perms, name="n_perms")
_assert_connectivity_key(adata, connectivity_key)
if genes is None:
if "highly_variable" in adata.var.columns:
genes = adata[:, adata.var.highly_variable.values].var_names.values
else:
genes = adata.var_names.values
genes = _assert_non_empty_sequence(genes, name="genes")
n_jobs = _get_n_cores(n_jobs)
start = logg.info(
f"Calculating for `{len(genes)}` genes using `{n_jobs}` core(s)")
w = _set_weight_class(adata, key=connectivity_key) # init weights
df = parallelize(
_moran_helper,
collection=genes,
extractor=pd.concat,
use_ixs=True,
n_jobs=n_jobs,
backend=backend,
show_progress_bar=show_progress_bar,
)(adata=adata,
weights=w,
transformation=transformation,
permutations=n_perms,
layer=layer,
seed=seed)
if corr_method is not None:
_, pvals_adj, _, _ = multipletests(df["pval_sim"].values,
alpha=0.05,
method=corr_method)
df[f"pval_sim_{corr_method}"] = pvals_adj
df.sort_values(by="I", ascending=False, inplace=True)
if copy:
logg.info("Finish", time=start)
return df
_save_data(adata, attr="uns", key="moranI", data=df, time=start)
x, y = np.triu_indices_from(np.empty((n_splits, n_splits)))
idx_splits = [(i, j) for i, j in zip(x, y)]
n_jobs = _get_n_cores(n_jobs)
start = logg.info(f"Calculating co-occurrence probabilities for\
`{len(interval)}` intervals\
`{len(idx_splits)}` split combinations\
using `{n_jobs}` core(s)")
out_lst = parallelize(
_co_occurrence_helper,
collection=idx_splits,
extractor=chain.from_iterable,
n_jobs=n_jobs,
backend=backend,
show_progress_bar=show_progress_bar,
)(
spatial_splits=spatial_splits,
labs_splits=labs_splits,
labs_unique=labs_unique,
interval=interval,
)
if len(idx_splits) == 1:
out = list(out_lst)[0]
else:
out = sum(list(out_lst)) / len(idx_splits)
if copy:
logg.info("Finish", time=start)
return out, interval
def spatial_autocorr(
adata: AnnData,
connectivity_key: str = Key.obsp.spatial_conn(),
genes: Optional[Union[str, Sequence[str]]] = None,
mode: Literal[
"moran",
"geary"] = SpatialAutocorr.MORAN.s, # type: ignore[assignment]
transformation: bool = True,
n_perms: Optional[int] = None,
two_tailed: bool = False,
corr_method: Optional[str] = "fdr_bh",
layer: Optional[str] = None,
seed: Optional[int] = None,
use_raw: bool = False,
copy: bool = False,
n_jobs: Optional[int] = None,
backend: str = "loky",
show_progress_bar: bool = True,
) -> Optional[pd.DataFrame]:
"""
Calculate Global Autocorrelation Statistic (Moran’s I or Geary's C).
See :cite:`pysal` for reference.
Parameters
----------
%(adata)s
%(conn_key)s
genes
List of gene names, as stored in :attr:`anndata.AnnData.var_names`, used to compute global
spatial autocorrelation statistic.
If `None`, it's computed :attr:`anndata.AnnData.var` ``['highly_variable']``, if present. Otherwise,
it's computed for all genes.
mode
Mode of score calculation:
- `{sp.MORAN.s!r}` - `Moran's I autocorrelation <https://en.wikipedia.org/wiki/Moran%27s_I>`_.
- `{sp.GEARY.s!r}` - `Geary's C autocorrelation <https://en.wikipedia.org/wiki/Geary%27s_C>`_.
transformation
If `True`, weights in :attr:`anndata.AnnData.obsp` ``['{key}']`` are row-normalized,
advised for analytic p-value calculation.
%(n_perms)s
If `None`, only p-values under normality assumption are computed.
two_tailed
If `True`, p-values are two-tailed, otherwise they are one-tailed.
%(corr_method)s
layer
Layer in :attr:`anndata.AnnData.layers` to use. If `None`, use :attr:`anndata.AnnData.X`.
%(seed)s
%(copy)s
%(parallelize)s
Returns
-------
If ``copy = True``, returns a :class:`pandas.DataFrame` with the following keys:
- `'I' or 'C'` - Moran's I or Geary's C statistic.
- `'pval_norm'` - p-value under normality assumption.
- `'var_norm'` - variance of `'score'` under normality assumption.
- `'{{p_val}}_{{corr_method}}'` - the corrected p-values if ``corr_method != None`` .
If ``n_perms != None`` is not None, additionally returns the following columns:
- `'pval_z_sim'` - p-value based on standard normal approximation from permutations.
- `'pval_sim'` - p-value based on permutations.
- `'var_sim'` - variance of `'score'` from permutations.
Otherwise, modifies the ``adata`` with the following key:
- :attr:`anndata.AnnData.uns` ``['moranI']`` - the above mentioned dataframe, if ``mode = {sp.MORAN.s!r}``.
- :attr:`anndata.AnnData.uns` ``['gearyC']`` - the above mentioned dataframe, if ``mode = {sp.GEARY.s!r}``.
"""
_assert_connectivity_key(adata, connectivity_key)
if genes is None:
if "highly_variable" in adata.var.columns:
genes = adata[:, adata.var.highly_variable.values].var_names.values
else:
genes = adata.var_names.values
genes = _assert_non_empty_sequence(genes, name="genes")
mode = SpatialAutocorr(mode) # type: ignore[assignment]
if TYPE_CHECKING:
assert isinstance(mode, SpatialAutocorr)
params = {
"mode": mode.s,
"transformation": transformation,
"two_tailed": two_tailed
}
if mode == SpatialAutocorr.MORAN:
params["func"] = _morans_i
params["stat"] = "I"
params["expected"] = -1.0 / (adata.shape[0] - 1) # expected score
params["ascending"] = False
elif mode == SpatialAutocorr.GEARY:
params["func"] = _gearys_c
params["stat"] = "C"
params["expected"] = 1.0
params["ascending"] = True
else:
raise NotImplementedError(f"Mode `{mode}` is not yet implemented.")
n_jobs = _get_n_cores(n_jobs)
vals = _get_obs_rep(adata[:, genes], use_raw=use_raw, layer=layer).T
g = adata.obsp[connectivity_key].copy()
# row-normalize
if transformation:
normalize(g, norm="l1", axis=1, copy=False)
score = params["func"](g, vals)
start = logg.info(
f"Calculating {mode}'s statistic for `{n_perms}` permutations using `{n_jobs}` core(s)"
)
if n_perms is not None:
_assert_positive(n_perms, name="n_perms")
perms = np.arange(n_perms)
score_perms = parallelize(
_score_helper,
collection=perms,
extractor=np.concatenate,
use_ixs=True,
n_jobs=n_jobs,
backend=backend,
show_progress_bar=show_progress_bar,
)(mode=mode, g=g, vals=vals, seed=seed)
else:
score_perms = None
with np.errstate(divide="ignore"):
pval_results = _p_value_calc(score, score_perms, g, params)
results = {params["stat"]: score}
results.update(pval_results)
df = pd.DataFrame(results, index=genes)
if corr_method is not None:
for pv in filter(lambda x: "pval" in x, df.columns):
_, pvals_adj, _, _ = multipletests(df[pv].values,
alpha=0.05,
method=corr_method)
df[f"{pv}_{corr_method}"] = pvals_adj
df.sort_values(by=params["stat"],
ascending=params["ascending"],
inplace=True)
if copy:
logg.info("Finish", time=start)
return df
_save_data(adata,
attr="uns",
key=params["mode"] + params["stat"],
data=df,
time=start)
def segment(
img: ImageContainer,
layer: Optional[str] = None,
method: Union[str, Callable[..., np.ndarray]] = "watershed",
channel: int = 0,
size: Optional[Union[int, Tuple[int, int]]] = None,
layer_added: Optional[str] = None,
copy: bool = False,
show_progress_bar: bool = True,
n_jobs: Optional[int] = None,
backend: str = "loky",
**kwargs: Any,
) -> Optional[ImageContainer]:
"""
Segment an image.
If ``size`` is defined, iterate over crops of that size and segment those. Recommended for large images.
Parameters
----------
%(img_container)s
%(img_layer)s
%(seg_blob.parameters)s
- `{m.WATERSHED.s!r}` - :func:`skimage.segmentation.watershed`.
%(custom_fn)s
channel
Channel index to use for segmentation.
%(size)s
%(layer_added)s
If `None`, use ``'segmented_{{model}}'``.
thresh
Threshold for creation of masked image. The areas to segment should be contained in this mask.
If `None`, it is determined by `Otsu's method <https://en.wikipedia.org/wiki/Otsu%27s_method>`_.
Only used if ``method = {m.WATERSHED.s!r}``.
geq
Treat ``thresh`` as upper or lower bound for defining areas to segment. If ``geq = True``, mask is defined
as ``mask = arr >= thresh``, meaning high values in ``arr`` denote areas to segment.
invert
Whether to segment an inverted array. Only used if ``method`` is one of :mod:`skimage` blob methods.
%(copy_cont)s
%(segment_kwargs)s
%(parallelize)s
kwargs
Keyword arguments for ``method``.
Returns
-------
If ``copy = True``, returns a new container with the segmented image in ``'{{layer_added}}'``.
Otherwise, modifies the ``img`` with the following key:
- :class:`squidpy.im.ImageContainer` ``['{{layer_added}}']`` - the segmented image.
"""
layer = img._get_layer(layer)
channel_dim = img[layer].dims[-1]
kind = SegmentationBackend.CUSTOM if callable(method) else SegmentationBackend(method)
layer_new = Key.img.segment(kind, layer_added=layer_added)
if kind in (SegmentationBackend.LOG, SegmentationBackend.DOG, SegmentationBackend.DOH):
segmentation_model: SegmentationModel = SegmentationBlob(model=kind)
elif kind == SegmentationBackend.WATERSHED:
segmentation_model = SegmentationWatershed()
elif kind == SegmentationBackend.CUSTOM:
if TYPE_CHECKING:
assert callable(method)
segmentation_model = SegmentationCustom(func=method)
else:
raise NotImplementedError(f"Model `{kind}` is not yet implemented.")
n_jobs = _get_n_cores(n_jobs)
crops: List[ImageContainer] = list(img.generate_equal_crops(size=size, as_array=False))
start = logg.info(f"Segmenting `{len(crops)}` crops using `{segmentation_model}` and `{n_jobs}` core(s)")
crops: List[ImageContainer] = parallelize( # type: ignore[no-redef]
_segment,
collection=crops,
unit="crop",
extractor=lambda res: list(chain.from_iterable(res)),
n_jobs=n_jobs,
backend=backend,
show_progress_bar=show_progress_bar and len(crops) > 1,
)(model=segmentation_model, layer=layer, layer_new=layer_new, channel=channel, **kwargs)
if isinstance(segmentation_model, SegmentationWatershed):
# By convention, segments are numbered from 1..number of segments within each crop.
# Next, we have to account for that before merging the crops so that segments are not confused.
# TODO use overlapping crops to not create confusion at boundaries
counter = 0
for crop in crops:
data = crop[layer_new].data
data[data > 0] += counter
counter += np.max(crop[layer_new].data)
res: ImageContainer = ImageContainer.uncrop(crops, shape=img.shape)
res._data = res.data.rename({channel_dim: f"{channel_dim}:{channel}"})
logg.info("Finish", time=start)
if copy:
return res
img.add_img(res, layer=layer_new, copy=False, channel_dim=res[layer_new].dims[-1])
def calculate_image_features(
adata: AnnData,
img: ImageContainer,
layer: Optional[str] = None,
features: Union[str, Sequence[str]] = ImageFeature.SUMMARY.s,
features_kwargs: Mapping[str, Mapping[str, Any]] = MappingProxyType({}),
key_added: str = "img_features",
copy: bool = False,
n_jobs: Optional[int] = None,
backend: str = "loky",
show_progress_bar: bool = True,
**kwargs: Any,
) -> Optional[pd.DataFrame]:
"""
Calculate image features for all observations in ``adata``.
Parameters
----------
%(adata)s
%(img_container)s
%(img_layer)s
features
Features to be calculated. Valid options are:
- `{f.TEXTURE.s!r}` - summary stats based on repeating patterns
:meth:`squidpy.im.ImageContainer.features_texture`.
- `{f.SUMMARY.s!r}` - summary stats of each image channel
:meth:`squidpy.im.ImageContainer.features_summary`.
- `{f.COLOR_HIST.s!r}` - counts in bins of image channel's histogram
:meth:`squidpy.im.ImageContainer.features_histogram`.
- `{f.SEGMENTATION.s!r}` - stats of a cell segmentation mask
:meth:`squidpy.im.ImageContainer.features_segmentation`.
- `{f.CUSTOM.s!r}` - extract features using a custom function
:meth:`squidpy.im.ImageContainer.features_custom`.
features_kwargs
Keyword arguments for the different features that should be generated, such as
``{{ {f.TEXTURE.s!r}: {{ ... }}, ... }}``.
key_added
Key in :attr:`anndata.AnnData.obsm` where to store the calculated features.
%(copy)s
%(parallelize)s
kwargs
Keyword arguments for :meth:`squidpy.im.ImageContainer.generate_spot_crops`.
Returns
-------
If ``copy = True``, returns a :class:`panda.DataFrame` where columns correspond to the calculated features.
Otherwise, modifies the ``adata`` object with the following key:
- :attr:`anndata.AnnData.uns` ``['{{key_added}}']`` - the above mentioned dataframe.
Raises
------
ValueError
If a feature is not known.
"""
layer = img._get_layer(layer)
if isinstance(features, (str, ImageFeature)):
features = [features]
features = sorted({ImageFeature(f).s for f in features})
n_jobs = _get_n_cores(n_jobs)
start = logg.info(
f"Calculating features `{list(features)}` using `{n_jobs}` core(s)")
res = parallelize(
_calculate_image_features_helper,
collection=adata.obs_names,
extractor=pd.concat,
n_jobs=n_jobs,
backend=backend,
show_progress_bar=show_progress_bar,
)(adata,
img,
layer=layer,
features=features,
features_kwargs=features_kwargs,
**kwargs)
if copy:
logg.info("Finish", time=start)
return res
_save_data(adata, attr="obsm", key=key_added, data=res, time=start)