-
Notifications
You must be signed in to change notification settings - Fork 19
/
mapping.py
executable file
·1891 lines (1552 loc) · 62.4 KB
/
mapping.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
from scipy.stats import binned_statistic
import hnswlib
import sklearn.utils.sparsefuncs as sf
import typing
from numba import njit, prange
import os
from os import path
import gc
from samalg import SAM
import time
from sklearn.preprocessing import StandardScaler
from . import q, ut, pd, sp, np, warnings, sc
from .analysis import _compute_csim
from .utils import prepend_var_prefix, to_vn, substr, sparse_knn, df_to_dict
from numba.core.errors import NumbaPerformanceWarning, NumbaWarning
warnings.filterwarnings("ignore", category=NumbaPerformanceWarning)
warnings.filterwarnings("ignore", category=NumbaWarning)
class SAMAP(object):
def __init__(
self,
sams: dict,
f_maps: typing.Optional[str] = "maps/",
names: typing.Optional[dict] = None,
keys: typing.Optional[dict] = None,
resolutions: typing.Optional[dict] = None,
gnnm: typing.Optional[tuple] = None,
save_processed: typing.Optional[bool] = True,
eval_thr: typing.Optional[float] = 1e-6
):
"""Initializes and preprocess data structures for SAMap algorithm.
Parameters
----------
sams : dict of string OR SAM
Dictionary of (indexed by species IDs):
The path to an unprocessed '.h5ad' `AnnData` object for organisms.
OR
A processed and already-run SAM object.
f_maps : string, optional, default 'maps/'
Path to the `maps` directory output by `map_genes.sh`.
By default assumes it is in the local directory.
names : dict of list of 2D tuples or Nx2 numpy.ndarray, optional, default None
If BLAST was run on a transcriptome with Fasta headers that do not match
the gene symbols used in the dataset, you can pass a list of tuples mapping
the Fasta header name to the Dataset gene symbol:
(Fasta header name , Dataset gene symbol). Transcripts with the same gene
symbol will be collapsed into a single node in the gene homology graph.
By default, the Fasta header IDs are assumed to be equivalent to the
gene symbols used in the dataset.
The above mapping should be contained in a dicitonary keyed by the corresponding species.
For example, if we have `hu` and `mo` species and the `hu` BLAST results need to be translated,
then `names = {'hu' : mapping}, where `mapping = [(Fasta header 1, Gene symbol 1), ... , (Fasta header n, Gene symbol n)]`.
keys : dict, optional, default None
Dictionary of obs keys indexed by species to use for determining maximum
neighborhood size of each cell.
resolutions : dict, optional, default None
Dictionary of leiden clustering resolutions indexed by species. This parameter is ignored if
`keys` is set.
gnnm : tuple(scipy.sparse.csr_matrix,numpy array, numpy array)
If the homology graph was already computed, you can pass it here in the form of a tuple:
(sparse adjacency matrix, species 1 genes, species 2 genes).
This is the tuple returned by `_calculate_blast_graph(...)` or `_coarsen_eggnog_graph(...)`.
save_processed : bool, optional, default False
If True saves the processed SAM objects corresponding to each species to an `.h5ad` file.
This argument is unused if preloaded SAM objects are passed in to SAMAP.
eval_thr : float, optional, default 1e-6
E-value threshold above which BLAST results will be filtered out.
"""
for key,data in zip(sams.keys(),sams.values()):
if not (isinstance(data, str) or isinstance(data, SAM)):
raise TypeError(f"Input data {key} must be either a path or a SAM object.")
ids = list(sams.keys())
if keys is None:
keys = {}
for sid in ids:
keys[sid] = 'leiden_clusters'
if resolutions is None:
resolutions = {}
for sid in ids:
resolutions[sid] = 3
for sid in ids:
data = sams[sid]
key = keys[sid]
res = resolutions[sid]
if isinstance(data, str):
print("Processing data {} from:\n{}".format(sid,data))
sam = SAM()
sam.load_data(data)
sam.preprocess_data(
sum_norm="cell_median",
norm="log",
thresh_low=0.0,
thresh_high=0.96,
min_expression=1,
)
sam.run(
preprocessing="StandardScaler",
npcs=100,
weight_PCs=False,
k=20,
n_genes=3000,
weight_mode='rms'
)
else:
sam = data
if key == "leiden_clusters":
sam.leiden_clustering(res=res)
if "PCs_SAMap" not in sam.adata.varm.keys():
prepare_SAMap_loadings(sam)
if save_processed and isinstance(data,str):
sam.save_anndata(data.split('.h5ad')[0]+'_pr.h5ad')
sams[sid] = sam
if gnnm is None:
gnnm, gns, gns_dict = _calculate_blast_graph(
ids, f_maps=f_maps, reciprocate=True, eval_thr=eval_thr
)
if names is not None:
gnnm, gns_dict, gns = _coarsen_blast_graph(
gnnm, gns, names
)
gnnm = _filter_gnnm(gnnm, thr=0.25)
else:
gnnm, gns, gns_dict = gnnm
gns_list=[]
ges_list=[]
for sid in ids:
prepend_var_prefix(sams[sid], sid)
ge = q(sams[sid].adata.var_names)
gn = gns_dict[sid]
gns_list.append(gn[np.in1d(gn, ge)])
ges_list.append(ge)
f = np.in1d(gns, np.concatenate(gns_list))
gns = gns[f]
gnnm = gnnm[f][:, f]
A = pd.DataFrame(data=np.arange(gns.size)[None, :], columns=gns)
ges = np.concatenate(ges_list)
ges = ges[np.in1d(ges, gns)]
ix = A[ges].values.flatten()
gnnm = gnnm[ix][:, ix]
gns = ges
gns_dict = {}
for i,sid in enumerate(ids):
gns_dict[sid] = ges[np.in1d(ges,gns_list[i])]
print(
"{} `{}` gene symbols match between the datasets and the BLAST graph.".format(
gns_dict[sid].size, sid
)
)
for sid in sams:
if not sp.sparse.issparse(sams[sid].adata.X):
sams[sid].adata.X = sp.sparse.csr_matrix(sams[sid].adata.X)
smap = _Samap_Iter(sams, gnnm, gns_dict, keys=keys)
self.sams = sams
self.gnnm = gnnm
self.gns_dict = gns_dict
self.gns = gns
self.ids = ids
self.smap = smap
def run(
self,
NUMITERS: typing.Optional[int] = 3,
NHS: typing.Optional[dict] = None,
crossK: typing.Optional[int] = 20,
N_GENE_CHUNKS: typing.Optional[int] = 1,
umap: typing.Optional[bool] = True,
ncpus: typing.Optional[float] = os.cpu_count(),
hom_edge_thr: typing.Optional[float] = 0,
hom_edge_mode: typing.Optional[str] = "pearson",
scale_edges_by_corr: typing.Optional[bool] = True,
neigh_from_keys: typing.Optional[dict] = None,
pairwise: typing.Optional[bool] = True
):
"""Runs the SAMap algorithm.
Parameters
----------
NUMITERS : int, optional, default 3
Runs SAMap for `NUMITERS` iterations.
NH1 : int, optional, default 3
Cells up to `NH1` hops away from a particular cell in organism 1
will be included in its neighborhood.
NH2 : int, optional, default 3
Cells up to `NH2` hops away from a particular cell in organism 2
will be included in its neighborhood.
crossK : int, optional, default 20
The number of cross-species edges to identify per cell.
N_GENE_CHUNKS : int, optional, default 1
When updating the edge weights in the BLAST homology graph, the operation
will be split up into `N_GENE_CHUNKS` chunks. For large datasets
(>50,000 cells), use more chunks (e.g. 4) to avoid running out of
memory.
umap : bool, optional, default True
If True, performs UMAP on the combined manifold to generate a 2D visualization.
If False, skips this step.
ncpus : int, optional, default `os.cpu_count()`
The number of CPUs to use when computing gene-gene correlations.
Defaults to using all available CPUs.
hom_edge_thr : float, optional, default 0
Edges with weight below `hom_edge_thr` in the homology graph will be set to zero.
hom_edge_mode: str, optional, default "pearson"
If "pearson", edge weights in the homology graph will be calculated using Pearson
correlation. If "mutual_info", edge weights will be calculated using normalized
mutual information. The latter requires package `fast-histogram` to be installed.
scale_edges_by_corr: bool, optional, default True
If True, rescale cell-cell cross-species edges by their expression similarities
(correlations).
neigh_from_key1 : bool, optional, default False
If True, species 1 neighborhoods are calculated directly from the chosen clustering (`self.key1`).
Cells within the same cluster belong to the same neighborhood.
neigh_from_key2 : bool, optional, default False
If True, species 2 neighborhoods are calculated directly from the chosen clustering (`self.key2`).
Cells within the same cluster belong to the same neighborhood.
pairwise: bool, optional, default True
If True, compute mutual nearest neighborhoods independently between each pair of species.
If False, compute mutual nearest neighborhoods between each species and all other species.
This parameter is ignored if there are only two species to be mapped.
`pairwise=True` would prevent outgroup species from being unmapped.
For example, if `pairwise=False`, when mapping three species like human, mouse, and zebrafish, human and mice
will preferentially map to each other, leaving zebrafish unmapped.
Set `pairwise=False` when you'd like to be able to tell which cell types are more or less similar
between different species. Set `pairwise=True` when you'd like to remove the effects of evolutionary
distance. This is particularly useful when using multiple reference species to annotate an unlabeled
dataset.
Returns
-------
samap - Species-merged SAM object
"""
self.pairwise = pairwise
ids = self.ids
sams = self.sams
gnnm = self.gnnm
gns_dict = self.gns_dict
gns = self.gns
smap = self.smap
if NHS is None:
NHS={}
for sid in ids:
NHS[sid] = 3
if neigh_from_keys is None:
neigh_from_keys={}
for sid in ids:
neigh_from_keys[sid] = False
start_time = time.time()
smap.run(
NUMITERS=NUMITERS,
NHS=NHS,
K=crossK,
NCLUSTERS=N_GENE_CHUNKS,
ncpus=ncpus,
THR=hom_edge_thr,
corr_mode=hom_edge_mode,
scale_edges_by_corr = scale_edges_by_corr,
neigh_from_keys=neigh_from_keys,
pairwise=pairwise
)
samap = smap.final_sam
self.samap = samap
self.ITER_DATA = smap.ITER_DATA
if umap:
print("Running UMAP on the stitched manifolds.")
sc.tl.umap(self.samap.adata,min_dist=0.1,init_pos='random',maxiter = 500 if self.samap.adata.shape[0] <= 10000 else 200)
ix = pd.Series(data = np.arange(samap.adata.shape[1]),index = samap.adata.var_names)[gns].values
rixer = pd.Series(index =np.arange(gns.size), data = ix)
try:
hom_graph = smap.GNNMS_corr[-1]
x,y = hom_graph.nonzero()
d = hom_graph.data
hom_graph = sp.sparse.coo_matrix((d,(rixer[x].values,rixer[y].values)),shape=(samap.adata.shape[1],)*2).tocsr()
samap.adata.varp["homology_graph_reweighted"] = hom_graph
self.gnnm_refined = hom_graph
except:
pass
x,y = gnnm.nonzero()
d = gnnm.data
gnnm = sp.sparse.coo_matrix((d,(rixer[x].values,rixer[y].values)),shape=(samap.adata.shape[1],)*2).tocsr()
samap.adata.varp["homology_graph"] = gnnm
samap.adata.uns["homology_gene_names_dict"] = gns_dict
self.gnnm = gnnm
self.gns = q(samap.adata.var_names)
gns_dict = {}
for sid in ids:
gns_dict[sid] = self.gns[np.in1d(self.gns,q(self.sams[sid].adata.var_names))]
self.gns_dict = gns_dict
if umap:
for sid in ids:
sams[sid].adata.obsm['X_umap_samap'] = self.samap.adata[sams[sid].adata.obs_names].obsm['X_umap']
self.run_time = time.time() - start_time
print("Elapsed time: {} minutes.".format(self.run_time / 60))
return samap
def run_umap(self):
print("Running UMAP on the stitched manifolds.")
ids = self.ids
sams = self.sams
sc.tl.umap(self.samap.adata,min_dist=0.1,init_pos='random', maxiter = 500 if self.samap.adata.shape[0] <= 10000 else 200)
for sid in ids:
sams[sid].adata.obsm['X_umap_samap'] = self.samap.adata[sams[sid].adata.obs_names].obsm['X_umap']
def plot_expression_overlap(self,gs,axes=None,#'#000098', COLOR2='#ffb900'
COLOR0='gray', COLORS=None, COLORC='#00ceb5',
s0 = 1, ss=None, sc = 10,
thr = 0.1,**kwargs):
"""Displays the expression overlap of two genes on the combined manifold.
Parameters
----------
gs : dict
Dictionary of genes to display, keyed by species IDs.
For example, human ('hu') and mouse ('ms') genes:
gs = {'hu':'TOP2A','ms':'Top2a'}
axes : matplotlib.pyplot.Axes, optional, default None
Displays the scatter plot on the provided axes if specified.
Otherwise creates a new figure.
COLOR0 : str, optional, default 'gray'
The color for cells that do not express `g1` or `g2`.
COLORS : dict, optional, default None
Dictionary of colors (hex codes) for cells expressing the
corresponding genes for each species. This dictionary is
keyed by species IDs. If not set, colors are chosen randomly.
COLORC : str, optional, default '#00ceb5'
The color for cells that overlap in
expression of the two genes.
s0 : int, optional, default 1
Marker size corresponding to `COLOR0`.
ss : dict, optional, default None
Dictionary of marker sizes corresponding to the colors in `COLORS`.
If not set, marker sizes default to 3.
sc : int, optional, default 10
Marker size corresponding to `COLORC`.
thr : float, optional, default 0.1
Threshold below which imputed expressions across species are zero'd out.
Keyword Arguments (**kwargs)
----------------------------
Most arguments accepted by matplotlib.pyplot.scatter are available.
Returns
-------
ax - matplotlib.pyplot.Axes
"""
if len(list(gs.keys()))<len(list(self.sams.keys())):
samap = SAM(counts = self.samap.adata[np.in1d(self.samap.adata.obs['species'],list(gs.keys()))])
else:
samap=self.samap
if ss is None:
ss={}
for sid in self.ids:
ss[sid] = 3
if COLORS is None:
COLORS={}
for sid in self.ids:
s = ''
for i in range(6):
s+=hex(np.random.randint(16))[-1].upper()
s='#'+s
COLORS[sid] = s
def hex_to_rgb(value):
value = value.lstrip('#')
lv = len(value)
lv = list(int(value[i:i + lv // 3], 16) for i in range(0, lv, lv // 3))
lv = [x/255 for x in lv]
return lv
nnm = samap.adata.obsp['connectivities']
su = nnm.sum(1).A.flatten()[:,None]
su[su==0]=1
nnm = nnm.multiply(1/su).tocsr()
AS={}
for sid in gs.keys():
g = gs[sid]
try:
AS[sid] = self.sams[sid].adata[:,g].X.A.flatten()
except KeyError:
try:
AS[sid] = self.sams[sid].adata[:,sid+'_'+g].X.A.flatten()
except KeyError:
raise KeyError(f'Gene not found in species {sid}')
davgs={}
for sid in gs.keys():
d = np.zeros(samap.adata.shape[0])
d[samap.adata.obs['species']==sid] = AS[sid]
davg = nnm.dot(d).flatten()
davg[davg<thr]=0
davgs[sid] = davg
davg = np.vstack(list(davgs.values())).min(0)
ma = np.vstack(list(davgs.values())).max()
for sid in gs.keys():
if davgs[sid].max()>0:
davgs[sid] = davgs[sid]/davgs[sid].max()
if davg.max()>0:
davg = davg/davg.max()
cs={}
for sid in gs.keys():
c = hex_to_rgb(COLORS[sid])+[0.0]
cs[sid] = np.vstack([c]*davg.size)
cs[sid][:,-1] = davgs[sid]
cc = hex_to_rgb(COLORC)+[0.0]
cc = np.vstack([cc]*davg.size)
cc[:,-1] = davg
ax = samap.scatter(projection = 'X_umap', colorspec = COLOR0, axes=axes, s = s0)
for sid in gs.keys():
samap.scatter(projection = 'X_umap', c = cs[sid], axes = ax, s = ss[sid],colorbar=False,**kwargs)
samap.scatter(projection = 'X_umap', c = cc, axes = ax, s = sc,colorbar=False,**kwargs)
return ax
def query_gene_pairs(self,gene):
""" Get BLAST and correlation scores of all genes connected
to the query gene.
Preferrably, genes are prepended with their species IDs.
For example, "hu_SOX2" instead of "SOX2".
Returns: Dictionary with "blast" and "correlation" keys with
the BLAST and correlation scores respectively for the queried
gene.
"""
ids = self.ids
qgene = None
if (gene in self.gns):
qgene = gene
else:
for sid in ids:
if sid+'_'+gene in self.gns:
qgene = sid+'_'+gene
break
if qgene is None:
raise ValueError(f"Query gene {gene} not found in dataset.")
a = self.gnnm[self.gns==qgene]
b = self.gnnm_refined[self.gns==qgene]
i1 = self.gns[a.nonzero()[1]]
i2 = self.gns[b.nonzero()[1]]
d1 = a.data
d2 = b.data
a = pd.Series(index=i1,data=d1)
b = pd.Series(index=i2,data=d2)
return {"blast":a,"correlation":b}
def query_gene_pair(self,gene1,gene2):
""" Get BLAST and correlation score for a pair of genes.
Preferrably, genes are prepended with their species IDs.
For example, "hu_SOX2" instead of "SOX2".
Returns: Dictionary with "blast" and "correlation" keys with
the BLAST and correlation scores respectively for the queried
gene pair.
"""
ids = self.ids
qgene1 = None
if (gene1 in self.gns):
qgene1 = gene1
else:
for sid in ids:
if sid+'_'+gene1 in self.gns:
qgene1 = sid+'_'+gene1
break
if qgene1 is None:
raise ValueError(f"Query gene {gene1} not found in dataset.")
qgene2 = None
if (gene2 in self.gns):
qgene2 = gene2
else:
for sid in ids:
if sid+'_'+gene2 in self.gns:
qgene2 = sid+'_'+gene2
break
if qgene2 is None:
raise ValueError(f"Query gene {gene2} not found in dataset.")
a = self.gnnm[self.gns==qgene1].A.flatten()[self.gns==qgene2][0]
b = self.gnnm_refined[self.gns==qgene1].A.flatten()[self.gns==qgene2][0]
return {"blast":a,"correlation":b}
def scatter(self,axes=None,COLORS=None,ss=None,**kwargs):
if ss is None:
ss={}
for sid in self.ids:
ss[sid] = 3
if COLORS is None:
COLORS={}
for sid in self.ids:
s = ''
for i in range(6):
s+=hex(np.random.randint(16))[-1].upper()
s='#'+s
COLORS[sid] = s
for sid in self.ids:
axes = self.sams[sid].scatter(projection = 'X_umap_samap', colorspec = COLORS[sid], axes = axes, s = ss[sid],colorbar=False,**kwargs)
return axes
def gui(self):
"""Launches a SAMGUI instance containing the two SAM objects."""
if 'SamapGui' not in self.__dict__:
try:
from samalg.gui import SAMGUI
except ImportError:
raise ImportError('Please install SAMGUI dependencies. See the README in the SAM github repository.')
sg = SAMGUI(sam = list(self.sams.values()), title = list(self.ids),default_proj='X_umap_samap')
self.SamapGui = sg
return sg.SamPlot
else:
return self.SamapGui.SamPlot
def refine_homology_graph(self, THR=0, NCLUSTERS=1, ncpus = os.cpu_count(), corr_mode='pearson', wscale = False):
gnnm = self.smap.refine_homology_graph(NCLUSTERS=NCLUSTERS, ncpus=ncpus, THR=THR, corr_mode=corr_mode, wscale=wscale)
samap = self.smap.samap
gns_dict = self.smap.gns_dict
gns = []
for sid in q(samap.adata.obs['species'])[np.sort(np.unique(samap.adata.obs['species'],return_index=True)[1])]:
gns.extend(gns_dict[sid])
gns=q(gns)
ix = pd.Series(data = np.arange(samap.adata.shape[1]),index = samap.adata.var_names)[gns].values
rixer = pd.Series(index =np.arange(gns.size), data = ix)
x,y = gnnm.nonzero()
d = gnnm.data
gnnm = sp.sparse.coo_matrix((d,(rixer[x].values,rixer[y].values)),shape=(samap.adata.shape[1],)*2).tocsr()
return gnnm
class _Samap_Iter(object):
def __init__(
self, sams, gnnm, gns_dict, keys=None
):
self.sams = sams
self.gnnm = gnnm
self.gnnmu = gnnm
self.gns_dict = gns_dict
if keys is None:
keys = {}
for sid in sams.keys():
keys[sid] = 'leiden_clusters'
self.keys = keys
self.GNNMS_corr = []
self.GNNMS_pruned = []
self.GNNMS_nnm = []
self.ITER_DATA = [
self.GNNMS_nnm,
self.GNNMS_corr,
self.GNNMS_pruned,
]
self.iter = 0
def refine_homology_graph(self, NCLUSTERS=1, ncpus=os.cpu_count(), THR=0, corr_mode='pearson', wscale=False):
if corr_mode=='mutual_info':
try:
from fast_histogram import histogram2d
except:
raise ImportError("Package `fast_histogram` must be installed for `corr_mode='mutual_info'`.");
sams = self.sams
gnnm = self.gnnm
gns_dict = self.gns_dict
gnnmu = self.gnnmu
keys = self.keys
sam4 = self.samap
gnnmu = _refine_corr(
sams,
sam4,
gnnm,
gns_dict,
THR=THR,
use_seq=False,
T1=0,
NCLUSTERS=NCLUSTERS,
ncpus=ncpus,
corr_mode=corr_mode,
wscale=wscale
)
return gnnmu
def run(self, NUMITERS=3, NHS=None, K=20, corr_mode='pearson', NCLUSTERS=1,
scale_edges_by_corr=True, THR=0, neigh_from_keys=None, pairwise=True,
ncpus=os.cpu_count()):
sams = self.sams
gnnm = self.gnnm
gns_dict = self.gns_dict
gnnmu = self.gnnmu
keys = self.keys
if NHS is None:
NHS={}
for sid in sams.keys():
NHS[sid] = 2
if neigh_from_keys is None:
neigh_from_keys={}
for sid in sams:
neigh_from_keys[sid] = False
gns = np.concatenate(list(gns_dict.values()))
if self.iter > 0:
sam4 = self.samap
for i in range(NUMITERS):
if self.iter > 0 and i == 0:
print("Calculating gene-gene correlations in the homology graph...")
gnnmu = self.refine_homology_graph(ncpus = ncpus, NCLUSTERS = NCLUSTERS, THR=THR, corr_mode=corr_mode)
self.GNNMS_corr.append(gnnmu)
self.gnnmu = gnnmu
gnnm2 = _get_pairs(sams, gnnmu, gns_dict, NOPs1=0, NOPs2=0)
self.GNNMS_pruned.append(gnnm2)
sam4 = _mapper(
sams,
gnnm2,
gns,
umap=False,
K=K,
NHS=NHS,
coarsen=True,
keys=keys,
scale_edges_by_corr=scale_edges_by_corr,
neigh_from_keys=neigh_from_keys,
pairwise=pairwise
)
sam4.adata.uns['mapping_K'] = K
self.samap = sam4
self.GNNMS_nnm.append(sam4.adata.obsp["connectivities"])
print("Iteration " + str(i+1) + " complete.")
print("Alignment scores:")
print(_avg_as(sam4))
self.iter += 1
if i < NUMITERS - 1:
print("Calculating gene-gene correlations in the homology graph...")
self.samap = sam4
gnnmu = self.refine_homology_graph(ncpus = ncpus, NCLUSTERS = NCLUSTERS, THR=THR, corr_mode=corr_mode)
self.GNNMS_corr.append(gnnmu)
self.gnnmu = gnnmu
gc.collect()
self.final_sam = sam4
def _avg_as(s):
x = q(s.adata.obs['species'])
xu = np.unique(x)
a = np.zeros((xu.size,xu.size))
for i in range(xu.size):
for j in range(xu.size):
if i!=j:
a[i,j] = s.adata.obsp['connectivities'][x==xu[i],:][:,x==xu[j]].sum(1).A.flatten().mean() / s.adata.uns['mapping_K']
return pd.DataFrame(data=a,index=xu,columns=xu)
@njit(parallel=True)
def _replace(X,xi,yi):
data = np.zeros(xi.size)
for i in prange(xi.size):
x=X[xi[i]]
y=X[yi[i]]
data[i] = ((x-x.mean())*(y-y.mean()) / x.std() / y.std()).sum() / x.size
return data
def _generate_coclustering_matrix(cl):
cl = ut.convert_annotations(np.array(list(cl)))
clu,cluc=np.unique(cl,return_counts=True)
v = np.zeros((cl.size,clu.size))
v[np.arange(v.shape[0]),cl]=1
v = sp.sparse.csr_matrix(v)
return v
def _mapper(
sams,
gnnm=None,
gn=None,
NHS=None,
umap=False,
mdata=None,
k=None,
K=20,
chunksize=20000,
coarsen=True,
keys=None,
scale_edges_by_corr=False,
neigh_from_keys=None,
pairwise=True,
**kwargs
):
if NHS is None:
NHS={}
for sid in sams.keys():
NHS[sid] = 3
if neigh_from_keys is None:
neigh_from_keys={}
for sid in sams.keys():
neigh_from_keys[sid] = False
if mdata is None:
mdata = _mapping_window(sams, gnnm, gn, K=K, pairwise=pairwise)
k1 = K
if keys is None:
keys = {}
for sid in sams.keys():
keys[sid] = 'leiden_clusters'
nnms_in={}
nnms_in0={}
flag=False
species_indexer=[]
for sid in sams.keys():
print(f"Expanding neighbourhoods of species {sid}...")
cl = sams[sid].get_labels(keys[sid])
_, ix, cluc = np.unique(cl, return_counts=True, return_inverse=True)
K = cluc[ix]
nnms_in0[sid] = sams[sid].adata.obsp["connectivities"].copy()
species_indexer.append(np.arange(sams[sid].adata.shape[0]))
if not neigh_from_keys[sid]:
nnm_in = _smart_expand(nnms_in0[sid], K, NH=NHS[sid])
nnm_in.data[:] = 1
nnms_in[sid]=nnm_in
else:
nnms_in[sid]=_generate_coclustering_matrix(cl)
flag=True
for i in range(1,len(species_indexer)):
species_indexer[i] += species_indexer[i-1].max()+1
if not flag:
nnm_internal = sp.sparse.block_diag(list(nnms_in.values())).tocsr()
nnm_internal0 = sp.sparse.block_diag(list(nnms_in0.values())).tocsr()
ovt = mdata["knn"]
ovt0 = ovt.copy()
ovt0.data[:]=1
B = ovt
# already sum-normalized per species
#s = B.sum(1).A
#s[s == 0] = 1
#B = B.multiply(1 / s).tocsr()
print("Indegree coarsening")
numiter = nnm_internal0.shape[0] // chunksize + 1
D = sp.sparse.csr_matrix((0, nnm_internal0.shape[0]))
if flag:
Cs=[]
for it,sid in enumerate(sams.keys()):
nfk = neigh_from_keys[sid]
if nfk:
Cs.append(nnms_in[sid].dot(nnms_in[sid].T.dot(B.T[species_indexer[it]])))
else:
Cs.append(nnms_in[sid].dot(B.T[species_indexer[it]]))
D = sp.sparse.vstack(Cs).T
del Cs
gc.collect()
else:
for bl in range(numiter):
print(str(bl) + "/" + str(numiter), D.shape)
C = B[bl * chunksize : (bl + 1) * chunksize].dot(nnm_internal.T)
C.data[C.data < 0.1] = 0
C.eliminate_zeros()
D = sp.sparse.vstack((D, C))
del C
gc.collect()
D = D.multiply(D.T).tocsr()
D.data[:] = D.data**0.5
mdata["xsim"] = D
if scale_edges_by_corr:
print('Rescaling edge weights by expression correlations.')
x,y = D.nonzero()
vals = _replace(mdata["wPCA"],x,y)
vals[vals<1e-3]=1e-3
# make a copy and write correlations
F = D.copy()
F.data[:] = vals
# normalize by maximum and rescale with tanh
ma = F.max(1).A
ma[ma==0]=1
F = F.multiply(1/ma).tocsr()
F.data[:] = _tanh_scale(F.data,center=0.7,scale=10)
# get max aligment score from before
ma = D.max(1).A
ma[ma==0]=1
# geometric mean expression correlation scores by alignment scores
D = F.multiply(D).tocsr()
D.data[:] = np.sqrt(D.data)
# get new max scores
ma2 = D.max(1).A
ma2[ma2==0]=1
# change new max scores to old max scores
D = D.multiply(ma/ma2).tocsr()
species_list = []
for sid in sams.keys():
species_list += [sid]*sams[sid].adata.shape[0]
species_list = np.array(species_list)
if not pairwise or len(sams.keys())==2:
Dk = sparse_knn(D, k1).tocsr()
denom = k1
else:
Dk=[]
for sid1 in sams.keys():
row=[]
for sid2 in sams.keys():
if sid1 != sid2:
Dsubk = sparse_knn(D[species_list==sid1][:,species_list==sid2], k1).tocsr()
else:
Dsubk = sp.sparse.csr_matrix((sams[sid1].adata.shape[0],)*2)
row.append(Dsubk)
Dk.append(sp.sparse.hstack(row))
Dk = sp.sparse.vstack(Dk).tocsr()
denom = (k1 * (len(sams.keys())-1))
sr = Dk.sum(1).A
x = 1 - sr.flatten() / denom
sr[sr==0]=1
st = Dk.sum(0).A.flatten()[None,:]
st[st==0]=1
proj = Dk.multiply(1 / sr).dot(Dk.multiply(1 / st)).tocsr()
z = proj.copy()
z.data[:] = 1
idx = np.where(z.sum(1).A.flatten() >= k1)[0]
omp = nnm_internal0
omp.data[:]=1
s = proj.max(1).A
s[s == 0] = 1
proj = proj.multiply(1 / s).tocsr()
X, Y = omp.nonzero()
X2 = X[np.in1d(X, idx)]
Y2 = Y[np.in1d(X, idx)]
omp = omp.tolil()
omp[X2, Y2] = np.vstack((proj[X2, Y2].A.flatten(), np.ones(X2.size) * 0.3)).max(
0
)
omp = nnm_internal0.tocsr()
NNM = omp.multiply(x[:, None])
NNM = (NNM+Dk).tolil()
NNM.setdiag(0)
print("Concatenating SAM objects...")
sam3 = _concatenate_sam(sams, NNM)
sam3.adata.obs["species"] = pd.Categorical(species_list)
sam3.adata.uns["gnnm_corr"] = mdata.get("gnnm_corr",None)
if umap:
print("Computing UMAP projection...")
sc.tl.umap(sam3.adata, min_dist=0.1, maxiter = 500 if sam3.adata.shape[0] <= 10000 else 200)
return sam3
def _refine_corr(
sams,
st,
gnnm,
gns_dict,
corr_mode="mutual_info",
THR=0,
use_seq=False,
T1=0.25,
NCLUSTERS=1,
ncpus=os.cpu_count(),
wscale=False
):
# import networkx as nx
gns = np.concatenate(list(gns_dict.values()))
x, y = gnnm.nonzero()
sam = list(sams.values())[0]
cl = sam.leiden_clustering(gnnm, res=0.5)
ix = np.argsort(cl)
NGPC = gns.size // NCLUSTERS + 1
ixs = []
for i in range(NCLUSTERS):
ixs.append(np.sort(ix[i * NGPC : (i + 1) * NGPC]))
assert np.concatenate(ixs).size == gns.size
GNNMSUBS = []
GNSUBS = []
for i in range(len(ixs)):
ixs[i] = np.unique(np.append(ixs[i], gnnm[ixs[i], :].nonzero()[1]))
gnnm_sub = gnnm[ixs[i], :][:, ixs[i]]
gnsub = gns[ixs[i]]