scFates 1.2.7__tar.gz → 1.2.8__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {scfates-1.2.7/scFates.egg-info → scfates-1.2.8}/PKG-INFO +1 -1
- {scfates-1.2.7 → scfates-1.2.8}/scFates/preprocessing/diffusion.py +4 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/preprocessing/pagoda2.py +4 -2
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tests/test_w_plots.py +56 -0
- scfates-1.2.8/scFates/tools/cluster.py +92 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/covariate.py +10 -3
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/root.py +4 -7
- {scfates-1.2.7 → scfates-1.2.8/scFates.egg-info}/PKG-INFO +1 -1
- scfates-1.2.7/scFates/tools/cluster.py +0 -121
- {scfates-1.2.7 → scfates-1.2.8}/LICENSE +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/MANIFEST.in +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/README.md +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/pyproject.toml +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/__init__.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/datasets/__init__.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/datasets/_datasets.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/datasets/test.h5ad +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/get/__init__.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/get/get.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/logging.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/pl.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/__init__.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/binned_pseudotime_meta.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/covariate.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/dendrogram.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/features.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/linearity_deviation.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/matrix.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/milestones.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/module_inclusion.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/modules.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/palette_tools.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/palettes.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/slide_cors.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/synchro_path.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/test_association.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/test_fork.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/trajectory.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/plot/utils.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/pp.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/preprocessing/__init__.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/settings.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tl.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/__init__.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/_test_monocle3.R +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/bifurcation_tools.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/conversion.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/correlation_tools.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/dendrogram.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/fit.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/graph_fitting.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/graph_operations.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/linearity_deviation.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/pseudotime.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/slide_cors.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/test_association.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates/tools/utils.py +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates.egg-info/SOURCES.txt +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates.egg-info/dependency_links.txt +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates.egg-info/requires.txt +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/scFates.egg-info/top_level.txt +0 -0
- {scfates-1.2.7 → scfates-1.2.8}/setup.cfg +0 -0
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: scFates
|
|
3
|
-
Version: 1.2.
|
|
3
|
+
Version: 1.2.8
|
|
4
4
|
Summary: scanpy compatible python suite for fast tree inference and advanced pseudotime downstream analysis
|
|
5
5
|
Author: Louis Faure
|
|
6
6
|
Project-URL: Repository, https://github.com/LouisFaure/scFates
|
|
@@ -62,6 +62,8 @@ def diffusion(
|
|
|
62
62
|
dict containing results from Palantir.
|
|
63
63
|
"""
|
|
64
64
|
|
|
65
|
+
adata = adata.copy() if copy else adata
|
|
66
|
+
|
|
65
67
|
logg.info("Running Diffusion maps ", reset=True)
|
|
66
68
|
|
|
67
69
|
data_df = pd.DataFrame(adata.obsm["X_pca"], index=adata.obs_names)
|
|
@@ -174,3 +176,5 @@ def diffusion(
|
|
|
174
176
|
|
|
175
177
|
logg.info(" finished", time=True, end=" " if settings.verbosity > 2 else "\n")
|
|
176
178
|
logg.hint("added \n" + logstr + uns_str)
|
|
179
|
+
|
|
180
|
+
return adata if copy else None
|
|
@@ -200,7 +200,9 @@ def batch_correct(
|
|
|
200
200
|
gene_av = (np.array(X.sum(axis=0)).ravel() + len(batches)) / (
|
|
201
201
|
depth.sum() + len(batches)
|
|
202
202
|
)
|
|
203
|
-
tc = np.vstack(
|
|
203
|
+
tc = np.vstack(
|
|
204
|
+
[X[(adata.obs[batch_key] == b).values, :].sum(axis=0) for b in batches]
|
|
205
|
+
)
|
|
204
206
|
tc = np.log(tc + 1) - np.log(
|
|
205
207
|
np.array([depth[adata.obs[batch_key].values == b].sum() for b in batches])
|
|
206
208
|
+ 1
|
|
@@ -211,7 +213,7 @@ def batch_correct(
|
|
|
211
213
|
|
|
212
214
|
batch = adata.obs[batch_key].cat.rename_categories(range(nbatches))
|
|
213
215
|
count_gene = np.repeat(np.arange(X.shape[0]), np.diff(X.indptr))
|
|
214
|
-
acc = np.transpose(np.vstack([count_gene, batch[X.indices]
|
|
216
|
+
acc = np.transpose(np.vstack([count_gene, batch.cat.codes.values[X.indices]]))
|
|
215
217
|
X.data = X.data / bc.values[acc[:, 0], acc[:, 1]]
|
|
216
218
|
logg.info(" depth scaling")
|
|
217
219
|
X = X.transpose()
|
|
@@ -86,6 +86,11 @@ def test_pipeline():
|
|
|
86
86
|
|
|
87
87
|
scf.tl.pseudotime(adata_2)
|
|
88
88
|
scf.tl.pseudotime(adata)
|
|
89
|
+
# automatic root on tips with all-negative values must still pick a tip
|
|
90
|
+
adata.obs["neg_t"] = -adata.obs.t - 1
|
|
91
|
+
adata_r = scf.tl.root(adata, "neg_t", tips_only=True, copy=True)
|
|
92
|
+
assert adata_r.uns["graph"]["root"] == adata.uns["graph"]["root"]
|
|
93
|
+
del adata.obs["neg_t"]
|
|
89
94
|
scf.tl.dendrogram(adata)
|
|
90
95
|
ax=scf.pl.dendrogram(adata, color_milestones=True,show=False)
|
|
91
96
|
|
|
@@ -189,6 +194,9 @@ def test_pipeline():
|
|
|
189
194
|
scf.tl.test_covariate(adata, "covariate")
|
|
190
195
|
scf.tl.test_covariate(adata, "covariate", trend_test=True)
|
|
191
196
|
scf.tl.test_covariate(adata, "covariate", nested=True)
|
|
197
|
+
# restricted to one segment of the tree
|
|
198
|
+
scf.tl.test_covariate(adata, "covariate", seg=str(adata.obs.seg.iloc[0]))
|
|
199
|
+
assert adata.var["cov_pval"].notna().all()
|
|
192
200
|
scf.pl.trend_covariate(
|
|
193
201
|
adata, adata.var_names[0], group_key="covariate", show_null=True
|
|
194
202
|
,show=False)
|
|
@@ -201,6 +209,14 @@ def test_pipeline():
|
|
|
201
209
|
scf.pl.trajectory(adata_2, root_milestone="A", milestones=["B"],show=False)
|
|
202
210
|
|
|
203
211
|
scf.tl.cluster(adata, n_neighbors=3, n_pcs=3)
|
|
212
|
+
# the requested layer is used, not .X: replacing .X by noise changes nothing
|
|
213
|
+
adata_n = adata.copy()
|
|
214
|
+
adata_n.X = np.random.default_rng(0).normal(size=adata.shape)
|
|
215
|
+
adata_n = scf.tl.cluster(adata_n, n_neighbors=3, n_pcs=3, copy=True)
|
|
216
|
+
adata_x = adata.copy()
|
|
217
|
+
adata_x.X = adata_x.layers["fitted"]
|
|
218
|
+
scf.tl.cluster(adata_x, layer="X", n_neighbors=3, n_pcs=3)
|
|
219
|
+
assert (adata_n.var.clusters.values == adata_x.var.clusters.values).all()
|
|
204
220
|
|
|
205
221
|
g = scf.pl.trends(adata, features=adata.var_names, return_genes=True)
|
|
206
222
|
scf.pl.trends(
|
|
@@ -400,3 +416,43 @@ def test_pipeline():
|
|
|
400
416
|
[-0.31524637, -0.31332134, -0.3392098, -0.33832952, -0.31277207],
|
|
401
417
|
rtol=1e-2,
|
|
402
418
|
)
|
|
419
|
+
|
|
420
|
+
|
|
421
|
+
def test_preprocessing(monkeypatch):
|
|
422
|
+
import sys
|
|
423
|
+
import types
|
|
424
|
+
import anndata as ad
|
|
425
|
+
from scipy.sparse import csr_matrix
|
|
426
|
+
|
|
427
|
+
rng = np.random.default_rng(0)
|
|
428
|
+
counts = rng.poisson(rng.gamma(0.5, 2, size=(1, 300)), size=(400, 300))
|
|
429
|
+
adata = ad.AnnData(csr_matrix(counts.astype(np.float32)))
|
|
430
|
+
adata.obs["batch"] = ["a"] * 200 + ["b"] * 200
|
|
431
|
+
|
|
432
|
+
scf.pp.filter_cells(adata, subset=False)
|
|
433
|
+
assert "outlier" in adata.obs
|
|
434
|
+
|
|
435
|
+
X_bc = scf.pp.batch_correct(adata, "batch", inplace=False)
|
|
436
|
+
assert X_bc.shape == adata.shape
|
|
437
|
+
|
|
438
|
+
# stand-in for palantir (not a test dependency) to check copy behaviour
|
|
439
|
+
utils = types.ModuleType("palantir.utils")
|
|
440
|
+
|
|
441
|
+
def run_diffusion_maps(data_df, n_components, knn, alpha):
|
|
442
|
+
ev = rng.normal(size=(data_df.shape[0], n_components))
|
|
443
|
+
return {
|
|
444
|
+
"EigenVectors": pd.DataFrame(ev, index=data_df.index),
|
|
445
|
+
"EigenValues": pd.Series(np.linspace(1, 0.1, n_components)),
|
|
446
|
+
}
|
|
447
|
+
|
|
448
|
+
utils.run_diffusion_maps = run_diffusion_maps
|
|
449
|
+
palantir = types.ModuleType("palantir")
|
|
450
|
+
palantir.utils = utils
|
|
451
|
+
monkeypatch.setitem(sys.modules, "palantir", palantir)
|
|
452
|
+
monkeypatch.setitem(sys.modules, "palantir.utils", utils)
|
|
453
|
+
|
|
454
|
+
adata = scf.datasets.test_adata()
|
|
455
|
+
del adata.obsm["X_diffusion"]
|
|
456
|
+
adata_d = scf.pp.diffusion(adata, n_components=5, multiscale=False, copy=True)
|
|
457
|
+
assert adata_d.obsm["X_diffusion"].shape == (adata.n_obs, 4)
|
|
458
|
+
assert "X_diffusion" not in adata.obsm
|
|
@@ -0,0 +1,92 @@
|
|
|
1
|
+
import pandas as pd
|
|
2
|
+
import scanpy as sc
|
|
3
|
+
from anndata import AnnData
|
|
4
|
+
|
|
5
|
+
from .. import logging as logg
|
|
6
|
+
from .. import settings
|
|
7
|
+
from .graph_fitting import get_data
|
|
8
|
+
|
|
9
|
+
|
|
10
|
+
def cluster(
|
|
11
|
+
adata: AnnData,
|
|
12
|
+
layer="fitted",
|
|
13
|
+
n_neighbors: int = 20,
|
|
14
|
+
n_pcs: int = 50,
|
|
15
|
+
metric: str = "cosine",
|
|
16
|
+
resolution: float = 1,
|
|
17
|
+
device: str = "cpu",
|
|
18
|
+
copy: bool = False,
|
|
19
|
+
):
|
|
20
|
+
"""\
|
|
21
|
+
Cluster features. Uses scanpy when using cpu, and rapids-singlecell when using gpu.
|
|
22
|
+
Dataset is transposed, PCA is calulcated and a nearest neighbor graph is generated
|
|
23
|
+
from PC space. Leiden algorithm is used for community detection.
|
|
24
|
+
|
|
25
|
+
|
|
26
|
+
Parameters
|
|
27
|
+
----------
|
|
28
|
+
adata
|
|
29
|
+
Annotated data matrix.
|
|
30
|
+
layer
|
|
31
|
+
Layer of feature to calculate clusters, by default fitted features
|
|
32
|
+
n_neighbors
|
|
33
|
+
Number of neighbors.
|
|
34
|
+
n_pcs
|
|
35
|
+
Number of PC to keep for PCA.
|
|
36
|
+
metric
|
|
37
|
+
distance metric to use for clustering.
|
|
38
|
+
resolution
|
|
39
|
+
Resolution parameter for leiden algorithm.
|
|
40
|
+
device
|
|
41
|
+
Run the analysis on 'cpu' (scanpy) or on 'gpu' (rapids-singlecell).
|
|
42
|
+
copy
|
|
43
|
+
Return a copy instead of writing to adata.
|
|
44
|
+
|
|
45
|
+
Returns
|
|
46
|
+
-------
|
|
47
|
+
adata : anndata.AnnData
|
|
48
|
+
if `copy=True` it returns or else add fields to `adata`:
|
|
49
|
+
|
|
50
|
+
`.var['clusters']`
|
|
51
|
+
module assignments for features.
|
|
52
|
+
"""
|
|
53
|
+
|
|
54
|
+
adata = adata.copy() if copy else adata
|
|
55
|
+
|
|
56
|
+
layer = "X" if layer not in adata.layers else layer
|
|
57
|
+
|
|
58
|
+
logg.info(
|
|
59
|
+
f"Clustering features using {layer} layer",
|
|
60
|
+
reset=True,
|
|
61
|
+
end="\n",
|
|
62
|
+
)
|
|
63
|
+
|
|
64
|
+
X, _ = get_data(adata, layer, None)
|
|
65
|
+
adata_s = AnnData(
|
|
66
|
+
X.values.T,
|
|
67
|
+
obs=pd.DataFrame(index=adata.var_names),
|
|
68
|
+
var=pd.DataFrame(index=adata.obs_names),
|
|
69
|
+
)
|
|
70
|
+
|
|
71
|
+
if device == "gpu":
|
|
72
|
+
try:
|
|
73
|
+
import rapids_singlecell as rsc
|
|
74
|
+
except ModuleNotFoundError:
|
|
75
|
+
raise Exception(
|
|
76
|
+
"rapids-singlecell is required for device='gpu', use device='cpu' instead!"
|
|
77
|
+
)
|
|
78
|
+
|
|
79
|
+
rsc.pp.pca(adata_s, n_comps=n_pcs)
|
|
80
|
+
rsc.pp.neighbors(adata_s, n_neighbors=n_neighbors, metric=metric)
|
|
81
|
+
rsc.tl.leiden(adata_s, resolution=resolution)
|
|
82
|
+
else:
|
|
83
|
+
sc.pp.pca(adata_s, n_comps=n_pcs)
|
|
84
|
+
sc.pp.neighbors(adata_s, n_neighbors=n_neighbors, metric=metric)
|
|
85
|
+
sc.tl.leiden(adata_s, resolution=resolution, flavor="igraph", n_iterations=2)
|
|
86
|
+
|
|
87
|
+
adata.var["clusters"] = adata_s.obs.leiden.values
|
|
88
|
+
|
|
89
|
+
logg.info(" finished", time=True, end=" " if settings.verbosity > 2 else "\n")
|
|
90
|
+
logg.hint("added \n" " .var['clusters'] identified modules.")
|
|
91
|
+
|
|
92
|
+
return adata if copy else None
|
|
@@ -128,10 +128,17 @@ def test_covariate(
|
|
|
128
128
|
logbase = np.log(adata.uns["log1p"]["base"])
|
|
129
129
|
else:
|
|
130
130
|
logbase = 1
|
|
131
|
-
|
|
132
|
-
|
|
133
|
-
|
|
131
|
+
# pseudotime and groups restricted to the same cells as the expression values
|
|
132
|
+
obs = adata.obs.loc[segcells]
|
|
133
|
+
groups = obs[group_key].cat.remove_unused_categories()
|
|
134
|
+
if len(groups.cat.categories) < 2:
|
|
135
|
+
raise ValueError(
|
|
136
|
+
f"At least two groups of '{group_key}' are needed"
|
|
137
|
+
+ (f" in segment {seg}" if seg is not None else "")
|
|
138
|
+
+ f", found: {list(groups.cat.categories)}."
|
|
134
139
|
)
|
|
140
|
+
dfs = [
|
|
141
|
+
pd.DataFrame(dict(t=obs.t, exp=X, groups=groups, logbase=logbase))
|
|
135
142
|
for X in Xgenes
|
|
136
143
|
]
|
|
137
144
|
|
|
@@ -84,15 +84,12 @@ def root(
|
|
|
84
84
|
)
|
|
85
85
|
avgs.loc[nodes] = avgs_temp
|
|
86
86
|
if tips_only:
|
|
87
|
+
# exclude non-tip nodes; NaN is skipped by idxmin/idxmax, unlike a
|
|
88
|
+
# placeholder value which could win when values are negative
|
|
87
89
|
mask = np.ones(avgs.shape, bool)
|
|
88
90
|
mask[adata.uns["graph"]["tips"]] = False
|
|
89
|
-
avgs[mask] =
|
|
90
|
-
if min_val
|
|
91
|
-
if tips_only:
|
|
92
|
-
avgs[mask] = avgs.max()
|
|
93
|
-
root = np.argmin(avgs)
|
|
94
|
-
else:
|
|
95
|
-
root = np.argmax(avgs)
|
|
91
|
+
avgs[mask] = np.nan
|
|
92
|
+
root = int(avgs.idxmin() if min_val else avgs.idxmax())
|
|
96
93
|
|
|
97
94
|
if circle:
|
|
98
95
|
d = 1e-6 + pairwise_distances(
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: scFates
|
|
3
|
-
Version: 1.2.
|
|
3
|
+
Version: 1.2.8
|
|
4
4
|
Summary: scanpy compatible python suite for fast tree inference and advanced pseudotime downstream analysis
|
|
5
5
|
Author: Louis Faure
|
|
6
6
|
Project-URL: Repository, https://github.com/LouisFaure/scFates
|
|
@@ -1,121 +0,0 @@
|
|
|
1
|
-
import scanpy as sc
|
|
2
|
-
from anndata import AnnData
|
|
3
|
-
from .. import logging as logg
|
|
4
|
-
from .. import settings
|
|
5
|
-
from .graph_fitting import get_data
|
|
6
|
-
|
|
7
|
-
|
|
8
|
-
def cluster(
|
|
9
|
-
adata: AnnData,
|
|
10
|
-
layer="fitted",
|
|
11
|
-
n_neighbors: int = 20,
|
|
12
|
-
n_pcs: int = 50,
|
|
13
|
-
metric: str = "cosine",
|
|
14
|
-
resolution: float = 1,
|
|
15
|
-
device: str = "cpu",
|
|
16
|
-
copy: bool = False,
|
|
17
|
-
):
|
|
18
|
-
"""\
|
|
19
|
-
Cluster features. Uses scanpy backend when using cpu, and cuml when using gpu.
|
|
20
|
-
Dataset is transposed, PCA is calulcated and a nearest neighbor graph is generated
|
|
21
|
-
from PC space. Leiden algorithm is used for community detection.
|
|
22
|
-
|
|
23
|
-
|
|
24
|
-
Parameters
|
|
25
|
-
----------
|
|
26
|
-
adata
|
|
27
|
-
Annotated data matrix.
|
|
28
|
-
layer
|
|
29
|
-
Layer of feature to calculate clusters, by default fitted features
|
|
30
|
-
n_neighbors
|
|
31
|
-
Number of neighbors.
|
|
32
|
-
n_pcs
|
|
33
|
-
Number of PC to keep for PCA.
|
|
34
|
-
metric
|
|
35
|
-
distance metric to use for clustering.
|
|
36
|
-
resolution
|
|
37
|
-
Resolution parameter for leiden algorithm.
|
|
38
|
-
device
|
|
39
|
-
run the analysis on 'cpu' with phenograph, or on 'gpu' with grapheno.
|
|
40
|
-
|
|
41
|
-
copy
|
|
42
|
-
Return a copy instead of writing to adata.
|
|
43
|
-
Returns
|
|
44
|
-
-------
|
|
45
|
-
adata : anndata.AnnData
|
|
46
|
-
if `copy=True` it returns subsetted or else subset (keeping only
|
|
47
|
-
significant features) and add fields to `adata`:
|
|
48
|
-
`.var['cluters']`
|
|
49
|
-
module assignments for features.
|
|
50
|
-
"""
|
|
51
|
-
|
|
52
|
-
layer = "X" if layer not in adata.layers else layer
|
|
53
|
-
|
|
54
|
-
logg.info(
|
|
55
|
-
f"Clustering features using {layer} layer",
|
|
56
|
-
reset=True,
|
|
57
|
-
end="\n",
|
|
58
|
-
)
|
|
59
|
-
|
|
60
|
-
X, use_rep = get_data(adata, layer, None)
|
|
61
|
-
|
|
62
|
-
adata_s = adata.copy().T
|
|
63
|
-
if device == "gpu":
|
|
64
|
-
from cuml.decomposition import PCA
|
|
65
|
-
|
|
66
|
-
pca = PCA(n_components=n_pcs)
|
|
67
|
-
adata_s.obsm["X_pca"] = pca.fit_transform(adata_s.X)
|
|
68
|
-
else:
|
|
69
|
-
sc.pp.pca(adata_s, n_comps=n_pcs)
|
|
70
|
-
|
|
71
|
-
sc.pp.neighbors(
|
|
72
|
-
adata_s,
|
|
73
|
-
n_neighbors=n_neighbors,
|
|
74
|
-
metric=metric,
|
|
75
|
-
method="rapids" if device == "gpu" else "umap",
|
|
76
|
-
)
|
|
77
|
-
|
|
78
|
-
if device == "gpu":
|
|
79
|
-
from scanpy.tools import _utils
|
|
80
|
-
|
|
81
|
-
adjacency = _utils._choose_graph(adata_s, None, None)
|
|
82
|
-
import cudf
|
|
83
|
-
import cugraph
|
|
84
|
-
from natsort import natsorted
|
|
85
|
-
import numpy as np
|
|
86
|
-
import pandas as pd
|
|
87
|
-
|
|
88
|
-
offsets = cudf.Series(adjacency.indptr)
|
|
89
|
-
indices = cudf.Series(adjacency.indices)
|
|
90
|
-
|
|
91
|
-
sources, targets = adjacency.nonzero()
|
|
92
|
-
weights = adjacency[sources, targets]
|
|
93
|
-
if isinstance(weights, np.matrix):
|
|
94
|
-
weights = np.asarray(weights).flatten()
|
|
95
|
-
weights = cudf.Series(weights)
|
|
96
|
-
|
|
97
|
-
g = cugraph.Graph()
|
|
98
|
-
|
|
99
|
-
if hasattr(g, "add_adj_list"):
|
|
100
|
-
g.add_adj_list(offsets, indices, weights)
|
|
101
|
-
else:
|
|
102
|
-
g.from_cudf_adjlist(offsets, indices, weights)
|
|
103
|
-
|
|
104
|
-
leiden_parts, _ = cugraph.leiden(g, resolution=resolution)
|
|
105
|
-
groups = (
|
|
106
|
-
leiden_parts.to_pandas()
|
|
107
|
-
.sort_values("vertex")[["partition"]]
|
|
108
|
-
.to_numpy()
|
|
109
|
-
.ravel()
|
|
110
|
-
)
|
|
111
|
-
adata_s.obs["leiden"] = pd.Categorical(
|
|
112
|
-
values=groups.astype("U"),
|
|
113
|
-
categories=natsorted(map(str, np.unique(groups))),
|
|
114
|
-
)
|
|
115
|
-
else:
|
|
116
|
-
sc.tl.leiden(adata_s, resolution=resolution, flavor="igraph", n_iterations=2)
|
|
117
|
-
|
|
118
|
-
adata.var["clusters"] = adata_s.obs.leiden
|
|
119
|
-
|
|
120
|
-
logg.info(" finished", time=True, end=" " if settings.verbosity > 2 else "\n")
|
|
121
|
-
logg.hint("added \n" " .var['clusters'] identified modules.")
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|