clone2vec 0.1.0__tar.gz → 0.1.1__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.
- {clone2vec-0.1.0 → clone2vec-0.1.1}/PKG-INFO +1 -1
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/__init__.py +1 -1
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/datasets.py +67 -5
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/plotting.py +74 -1
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/utils.py +107 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec.egg-info/PKG-INFO +1 -1
- {clone2vec-0.1.0 → clone2vec-0.1.1}/setup.cfg +1 -1
- {clone2vec-0.1.0 → clone2vec-0.1.1}/tests/test_associations.py +1 -1
- {clone2vec-0.1.0 → clone2vec-0.1.1}/tests/test_pl.py +12 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/tests/test_utils.py +18 -1
- {clone2vec-0.1.0 → clone2vec-0.1.1}/README.md +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/associations.py +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/catboost.py +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/eigenvalues.py +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/embeddings.py +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/integration.py +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/preprocessing.py +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/seurat.py +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec/tools.py +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec.egg-info/SOURCES.txt +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec.egg-info/dependency_links.txt +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec.egg-info/requires.txt +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/clone2vec.egg-info/top_level.txt +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/pyproject.toml +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/tests/test_embeddings.py +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/tests/test_integration.py +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/tests/test_pp.py +0 -0
- {clone2vec-0.1.0 → clone2vec-0.1.1}/tests/test_tl.py +0 -0
|
@@ -1,27 +1,89 @@
|
|
|
1
1
|
from __future__ import annotations
|
|
2
2
|
|
|
3
3
|
from scanpy import read, AnnData
|
|
4
|
+
from typing import Literal
|
|
4
5
|
from pathlib import Path
|
|
5
6
|
|
|
7
|
+
import scanpy as sc
|
|
8
|
+
|
|
6
9
|
zenodo_record_lt = "15334396"
|
|
10
|
+
zenodo_record_c2v = "19973466"
|
|
11
|
+
|
|
12
|
+
logg = sc.logging
|
|
13
|
+
|
|
14
|
+
__all__ = [
|
|
15
|
+
"Liu_NSCLC_CD8",
|
|
16
|
+
"Weinreb_in_vitro",
|
|
17
|
+
"Erickson_murine_development",
|
|
18
|
+
]
|
|
19
|
+
|
|
20
|
+
def __dir__():
|
|
21
|
+
return sorted(__all__)
|
|
22
|
+
|
|
23
|
+
def Liu_NSCLC_CD8(
|
|
24
|
+
embedding_type: Literal["gex", "c2v"] = "gex",
|
|
25
|
+
file_path: str | Path | None = None,
|
|
26
|
+
) -> AnnData:
|
|
27
|
+
"""
|
|
28
|
+
Dataset from Liu et al. [PMID: 35121991] with CD8 T cells from NSCLC.
|
|
29
|
+
|
|
30
|
+
Parameters
|
|
31
|
+
----------
|
|
32
|
+
embedding_type : Literal["gex", "c2v"], optional
|
|
33
|
+
Type of embedding to use, by default "gex".
|
|
34
|
+
file_path : str | Path | None, optional
|
|
35
|
+
Path where .h5ad-container will be stored, by default None.
|
|
36
|
+
|
|
37
|
+
Returns
|
|
38
|
+
-------
|
|
39
|
+
AnnData
|
|
40
|
+
Annotated data matrix with the dataset.
|
|
41
|
+
"""
|
|
42
|
+
if embedding_type == "c2v":
|
|
43
|
+
logg.info("using clonal embedding (compositions in adata.X)")
|
|
44
|
+
elif embedding_type == "gex":
|
|
45
|
+
logg.info("using gene expression embedding")
|
|
46
|
+
else:
|
|
47
|
+
raise ValueError("embedding_type must be one of 'gex' or 'c2v'")
|
|
48
|
+
|
|
49
|
+
if file_path is None:
|
|
50
|
+
file_path = f"data/Liu_CD8_{embedding_type}.h5ad"
|
|
51
|
+
|
|
52
|
+
url = f"https://zenodo.org/records/{zenodo_record_c2v}/files/Liu_CD8_{embedding_type}.h5ad"
|
|
53
|
+
|
|
54
|
+
adata = read(file_path, backup_url=url, sparse=True, cache=True)
|
|
55
|
+
return adata
|
|
7
56
|
|
|
8
57
|
def Weinreb_in_vitro(
|
|
9
|
-
|
|
58
|
+
embedding_type: Literal["gex", "c2v"] = "gex",
|
|
59
|
+
file_path: str | Path | None = None,
|
|
10
60
|
) -> AnnData:
|
|
11
61
|
"""
|
|
12
|
-
Dataset from [PMID: 31974159] with in vitro hematopoiesis.
|
|
62
|
+
Dataset from Weinreb et al. [PMID: 31974159] with in vitro hematopoiesis.
|
|
13
63
|
|
|
14
64
|
Parameters
|
|
15
65
|
----------
|
|
16
|
-
|
|
17
|
-
|
|
66
|
+
embedding_type : Literal["gex", "c2v"], optional
|
|
67
|
+
Type of embedding to use, by default "gex".
|
|
68
|
+
file_path : str | Path | None, optional
|
|
69
|
+
Path where .h5ad-container will be stored, by default None.
|
|
18
70
|
|
|
19
71
|
Returns
|
|
20
72
|
-------
|
|
21
73
|
AnnData
|
|
22
74
|
Annotated data matrix with the dataset.
|
|
23
75
|
"""
|
|
24
|
-
|
|
76
|
+
if embedding_type == "c2v":
|
|
77
|
+
logg.info("using clonal embedding (compositions in adata.X)")
|
|
78
|
+
elif embedding_type == "gex":
|
|
79
|
+
logg.info("using gene expression embedding")
|
|
80
|
+
else:
|
|
81
|
+
raise ValueError("embedding_type must be one of 'gex' or 'c2v'")
|
|
82
|
+
|
|
83
|
+
if file_path is None:
|
|
84
|
+
file_path = f"data/Weinreb_in_vitro_{embedding_type}.h5ad"
|
|
85
|
+
|
|
86
|
+
url = f"https://zenodo.org/records/{zenodo_record_c2v}/files/Weinreb_in_vitro_{embedding_type}.h5ad"
|
|
25
87
|
|
|
26
88
|
adata = read(file_path, backup_url=url, sparse=True, cache=True)
|
|
27
89
|
return adata
|
|
@@ -35,11 +35,84 @@ __all__ = [
|
|
|
35
35
|
"scaled_dotplot",
|
|
36
36
|
"scatter2vars",
|
|
37
37
|
"pca_loadings",
|
|
38
|
+
"embedding",
|
|
38
39
|
]
|
|
39
40
|
|
|
40
41
|
def __dir__():
|
|
41
42
|
return sorted(__all__)
|
|
42
43
|
|
|
44
|
+
def embedding(
|
|
45
|
+
adata: sc.AnnData,
|
|
46
|
+
obsm_name: str | None = None,
|
|
47
|
+
obsm_component: int | str | list[int] | list[str] | None = None,
|
|
48
|
+
basis: str = "X_umap",
|
|
49
|
+
**kwargs
|
|
50
|
+
) -> matplotlib.axes.Axes | matplotlib.figure.Figure | None:
|
|
51
|
+
"""
|
|
52
|
+
Wrapper around sc.pl.embedding that allows to plot embedding from obsm.
|
|
53
|
+
|
|
54
|
+
Parameters
|
|
55
|
+
----------
|
|
56
|
+
adata: sc.AnnData
|
|
57
|
+
Annotated data matrix.
|
|
58
|
+
obsm_name: str | None
|
|
59
|
+
Key for embedding in `adata.obsm`.
|
|
60
|
+
obsm_component: int | str | list[int] | list[str] | None
|
|
61
|
+
Component(s) to plot. Can be:
|
|
62
|
+
|
|
63
|
+
- ``int``: column index (works for both arrays and DataFrames).
|
|
64
|
+
- ``str``: column name (only when ``adata.obsm[obsm_name]`` is a
|
|
65
|
+
:class:`~pandas.DataFrame`).
|
|
66
|
+
- ``list[int]`` or ``list[str]``: multiple components; each is
|
|
67
|
+
plotted as a separate colour channel via *sc.pl.embedding*.
|
|
68
|
+
- ``None`` (default): uses component ``0``.
|
|
69
|
+
basis: str
|
|
70
|
+
Key for embedding coordinates in `adata.obsm`.
|
|
71
|
+
**kwargs
|
|
72
|
+
Additional arguments to pass to `sc.pl.embedding`.
|
|
73
|
+
|
|
74
|
+
Returns
|
|
75
|
+
-------
|
|
76
|
+
res: matplotlib.axes.Axes | matplotlib.figure.Figure | None
|
|
77
|
+
Axes object with the plot.
|
|
78
|
+
"""
|
|
79
|
+
tmp_keys: list[str] = [] # obs keys to clean up afterwards
|
|
80
|
+
if obsm_name:
|
|
81
|
+
if obsm_component is None:
|
|
82
|
+
obsm_component = 0
|
|
83
|
+
|
|
84
|
+
obsm_data = adata.obsm[obsm_name]
|
|
85
|
+
is_df = isinstance(obsm_data, pd.DataFrame)
|
|
86
|
+
|
|
87
|
+
# Normalise to a list so that the loop below works for both
|
|
88
|
+
# scalar and list inputs.
|
|
89
|
+
components = (
|
|
90
|
+
obsm_component
|
|
91
|
+
if isinstance(obsm_component, list)
|
|
92
|
+
else [obsm_component]
|
|
93
|
+
)
|
|
94
|
+
|
|
95
|
+
for comp in components:
|
|
96
|
+
obs_key = f"{obsm_name}_{comp}"
|
|
97
|
+
if is_df:
|
|
98
|
+
if isinstance(comp, str):
|
|
99
|
+
adata.obs[obs_key] = list(obsm_data[comp])
|
|
100
|
+
else:
|
|
101
|
+
adata.obs[obs_key] = list(obsm_data.iloc[:, comp])
|
|
102
|
+
else:
|
|
103
|
+
adata.obs[obs_key] = list(obsm_data[:, comp])
|
|
104
|
+
tmp_keys.append(obs_key)
|
|
105
|
+
|
|
106
|
+
kwargs["color"] = tmp_keys if len(tmp_keys) > 1 else tmp_keys[0]
|
|
107
|
+
|
|
108
|
+
if "ax" in kwargs:
|
|
109
|
+
kwargs["show"] = False
|
|
110
|
+
res = sc.pl.embedding(adata, basis=basis, **kwargs)
|
|
111
|
+
for key in tmp_keys:
|
|
112
|
+
del adata.obs[key]
|
|
113
|
+
return res
|
|
114
|
+
|
|
115
|
+
|
|
43
116
|
def pca_loadings(
|
|
44
117
|
adata: sc.AnnData,
|
|
45
118
|
key: str = "gPCs",
|
|
@@ -509,7 +582,7 @@ def group_scatter(
|
|
|
509
582
|
if kwargs_background is None:
|
|
510
583
|
kwargs_background = [{} for i in range(len(groups))]
|
|
511
584
|
if isinstance(kwargs_background, dict):
|
|
512
|
-
kwargs_background = [
|
|
585
|
+
kwargs_background = [kwargs_background.copy() for i in range(len(groups))]
|
|
513
586
|
elif len(kwargs_background) != len(groups):
|
|
514
587
|
raise ValueError("kwargs_background must be a dict or a list of dict with length equal to the number of groups")
|
|
515
588
|
if kwargs_group is None:
|
|
@@ -15,6 +15,8 @@ __all__ = [
|
|
|
15
15
|
"get_connectivity_matrix",
|
|
16
16
|
"gs",
|
|
17
17
|
"regress_categories",
|
|
18
|
+
"impute",
|
|
19
|
+
"laplacian_eigenmaps",
|
|
18
20
|
]
|
|
19
21
|
|
|
20
22
|
def __dir__():
|
|
@@ -507,6 +509,10 @@ def _nan_mask(X: np.ndarray | sp.sparse.csr_matrix) -> np.ndarray:
|
|
|
507
509
|
mask = np.isnan(arr).any(axis=1)
|
|
508
510
|
return mask
|
|
509
511
|
|
|
512
|
+
|
|
513
|
+
|
|
514
|
+
|
|
515
|
+
|
|
510
516
|
def gs(
|
|
511
517
|
adata: sc.AnnData,
|
|
512
518
|
use_rep: str = "X_pca",
|
|
@@ -615,4 +621,105 @@ def gs(
|
|
|
615
621
|
f" .obs['{obs_key}'] geometric sketching labels",
|
|
616
622
|
" .uns['gs'] parameters",
|
|
617
623
|
]
|
|
624
|
+
logg.info(" finished ({time_passed})", deep="\n".join(lines), time=start)
|
|
625
|
+
|
|
626
|
+
def laplacian_eigenmaps(
|
|
627
|
+
adata: sc.AnnData,
|
|
628
|
+
obsp_key: str = "connectivities",
|
|
629
|
+
n_components: int = 20,
|
|
630
|
+
key_added: str = "X_laplacian",
|
|
631
|
+
norm_laplacian: bool = True,
|
|
632
|
+
uns_key: str = "laplacian_eigenmaps",
|
|
633
|
+
) -> None:
|
|
634
|
+
"""
|
|
635
|
+
Compute Laplacian eigenmaps from a sparse graph stored in ``adata.obsp``.
|
|
636
|
+
|
|
637
|
+
The function builds the (normalised) graph Laplacian from the adjacency
|
|
638
|
+
matrix in ``adata.obsp[obsp_key]``, computes the smallest non-trivial
|
|
639
|
+
eigenvectors, and stores them in ``adata.obsm[key_added]``.
|
|
640
|
+
|
|
641
|
+
Parameters
|
|
642
|
+
----------
|
|
643
|
+
adata : sc.AnnData
|
|
644
|
+
Annotated data matrix. Must contain a sparse adjacency matrix in
|
|
645
|
+
``adata.obsp[obsp_key]``.
|
|
646
|
+
obsp_key : str, optional
|
|
647
|
+
Key in ``adata.obsp`` for the adjacency / connectivity matrix.
|
|
648
|
+
Default is ``"connectivities"``.
|
|
649
|
+
n_components : int, optional
|
|
650
|
+
Number of eigenvectors to compute (excluding the trivial constant
|
|
651
|
+
eigenvector). Default is 20.
|
|
652
|
+
key_added : str, optional
|
|
653
|
+
Key in ``adata.obsm`` where the embedding is stored.
|
|
654
|
+
Default is ``"X_laplacian"``.
|
|
655
|
+
norm_laplacian : bool, optional
|
|
656
|
+
If ``True``, use the symmetric normalised Laplacian
|
|
657
|
+
:math:`I - D^{-1/2} W D^{-1/2}`. Otherwise use the combinatorial
|
|
658
|
+
Laplacian :math:`D - W`. Default is ``True``.
|
|
659
|
+
uns_key : str, optional
|
|
660
|
+
Name of the new column in ``adata.uns`` to store the Laplacian eigenmaps parameters,
|
|
661
|
+
by default "laplacian_eigenmaps".
|
|
662
|
+
|
|
663
|
+
Returns
|
|
664
|
+
-------
|
|
665
|
+
None
|
|
666
|
+
The embedding is stored in place in ``adata.obsm[key_added]``.
|
|
667
|
+
"""
|
|
668
|
+
from scipy.sparse.linalg import eigsh
|
|
669
|
+
|
|
670
|
+
line_before = f"\n using .obsp['{obsp_key}'], n_components={n_components}"
|
|
671
|
+
start = logg.info("computing Laplacian eigenmaps", deep=line_before)
|
|
672
|
+
|
|
673
|
+
if obsp_key not in adata.obsp:
|
|
674
|
+
raise KeyError(f"'{obsp_key}' not found in adata.obsp")
|
|
675
|
+
|
|
676
|
+
W = adata.obsp[obsp_key]
|
|
677
|
+
if not sp.issparse(W):
|
|
678
|
+
W = sp.csr_matrix(W)
|
|
679
|
+
else:
|
|
680
|
+
W = W.tocsr()
|
|
681
|
+
|
|
682
|
+
if (W - W.T).nnz > 0:
|
|
683
|
+
W = (W + W.T) / 2.0
|
|
684
|
+
|
|
685
|
+
n = W.shape[0]
|
|
686
|
+
if n_components >= n:
|
|
687
|
+
raise ValueError(
|
|
688
|
+
f"n_components ({n_components}) must be less than the number of "
|
|
689
|
+
f"observations ({n})"
|
|
690
|
+
)
|
|
691
|
+
|
|
692
|
+
d = np.asarray(W.sum(axis=1)).ravel()
|
|
693
|
+
|
|
694
|
+
if norm_laplacian:
|
|
695
|
+
d_inv_sqrt = np.zeros_like(d)
|
|
696
|
+
mask = d > 0
|
|
697
|
+
d_inv_sqrt[mask] = 1.0 / np.sqrt(d[mask])
|
|
698
|
+
D_inv_sqrt = sp.diags(d_inv_sqrt)
|
|
699
|
+
L = sp.eye(n, format="csr") - D_inv_sqrt @ W @ D_inv_sqrt
|
|
700
|
+
else:
|
|
701
|
+
D = sp.diags(d)
|
|
702
|
+
L = D - W
|
|
703
|
+
|
|
704
|
+
eigenvalues, eigenvectors = eigsh(L, k=n_components + 1, which="SM")
|
|
705
|
+
|
|
706
|
+
order = np.argsort(eigenvalues)
|
|
707
|
+
eigenvalues = eigenvalues[order]
|
|
708
|
+
eigenvectors = eigenvectors[:, order]
|
|
709
|
+
|
|
710
|
+
embedding = eigenvectors[:, 1:]
|
|
711
|
+
|
|
712
|
+
adata.obsm[key_added] = embedding
|
|
713
|
+
adata.uns[uns_key] = {
|
|
714
|
+
"eigenvalues": eigenvalues[1:],
|
|
715
|
+
"components": n_components,
|
|
716
|
+
"norm_laplacian": norm_laplacian,
|
|
717
|
+
"key_added": key_added,
|
|
718
|
+
"obsp_key": obsp_key,
|
|
719
|
+
}
|
|
720
|
+
lines = [
|
|
721
|
+
"added",
|
|
722
|
+
f" .obsm['{key_added}'] Laplacian eigenmaps ({n_components} components)",
|
|
723
|
+
f" .uns['{uns_key}'] parameters",
|
|
724
|
+
]
|
|
618
725
|
logg.info(" finished ({time_passed})", deep="\n".join(lines), time=start)
|
|
@@ -340,7 +340,7 @@ def test_associations_binomial_gam(adata_binomial):
|
|
|
340
340
|
|
|
341
341
|
def test_weighted_corr_vs_manual():
|
|
342
342
|
"""Verify that _fast_corr with weights matches a manual weighted Pearson."""
|
|
343
|
-
from
|
|
343
|
+
from clone2vec.associations import _fast_corr
|
|
344
344
|
|
|
345
345
|
rng = np.random.RandomState(42)
|
|
346
346
|
n, p, q = 100, 5, 2
|
|
@@ -74,6 +74,18 @@ def test_group_kde_returns_figure(matrix_kind, adata_with_umap_dense, adata_with
|
|
|
74
74
|
fig = c2v.pl.group_kde(ad, groupby="cell_type", groups=group, basis="X_umap", return_fig=True)
|
|
75
75
|
assert isinstance(fig, plt.Figure)
|
|
76
76
|
|
|
77
|
+
# ===== embedding =====
|
|
78
|
+
|
|
79
|
+
def test_embedding_with_obsm(adata_with_umap_dense):
|
|
80
|
+
ad = adata_with_umap_dense.copy()
|
|
81
|
+
rng = np.random.RandomState(42)
|
|
82
|
+
ad.obsm["X_test_emb"] = rng.rand(ad.n_obs, 3)
|
|
83
|
+
ax = c2v.pl.embedding(ad, obsm_name="X_test_emb", obsm_component=1, basis="X_umap", show=False)
|
|
84
|
+
# temporary column should be removed after plotting
|
|
85
|
+
assert "X_test_emb_1" not in ad.obs.columns
|
|
86
|
+
plt.close("all")
|
|
87
|
+
|
|
88
|
+
|
|
77
89
|
# ===== loss_history and clone_size =====
|
|
78
90
|
|
|
79
91
|
def test_loss_history_plots_line(clones):
|
|
@@ -236,4 +236,21 @@ def test_impute_categorical(adata_dense):
|
|
|
236
236
|
assert "ct_with_na_imp" in ad.obs
|
|
237
237
|
imputed_vals = ad.obs["ct_with_na_imp"].astype(str).values
|
|
238
238
|
assert np.sum(imputed_vals == "NA") == 0
|
|
239
|
-
assert "impute_prob" in ad.obsm
|
|
239
|
+
assert "impute_prob" in ad.obsm
|
|
240
|
+
|
|
241
|
+
|
|
242
|
+
# ===== Tests for laplacian_eigenmaps =====
|
|
243
|
+
|
|
244
|
+
def test_laplacian_eigenmaps():
|
|
245
|
+
n, k = 50, 5
|
|
246
|
+
# ring graph
|
|
247
|
+
rows = list(range(n)) + list(range(1, n)) + [0]
|
|
248
|
+
cols = list(range(1, n)) + [0] + list(range(n))
|
|
249
|
+
data = [1.0] * len(rows)
|
|
250
|
+
W = sp.csr_matrix((data, (rows, cols)), shape=(n, n))
|
|
251
|
+
ad = sc.AnnData(X=np.zeros((n, 2)), obsp={"connectivities": W})
|
|
252
|
+
c2v.utils.laplacian_eigenmaps(ad, n_components=k)
|
|
253
|
+
assert "X_laplacian" in ad.obsm
|
|
254
|
+
assert ad.obsm["X_laplacian"].shape == (n, k)
|
|
255
|
+
assert "laplacian_eigenmaps" in ad.uns
|
|
256
|
+
assert len(ad.uns["laplacian_eigenmaps"]["eigenvalues"]) == k
|
|
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
|