-
Notifications
You must be signed in to change notification settings - Fork 4
/
scRNA_scATAC1.r
1977 lines (1765 loc) · 69.8 KB
/
scRNA_scATAC1.r
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
# package library and install
if (!requireNamespace("BiocManager", quietly = TRUE))
install.packages("BiocManager")
if (!requireNamespace("EnsDb.Hsapiens.v86", quietly = TRUE))
BiocManager::install("EnsDb.Hsapiens.v86")
if (!requireNamespace("EnsDb.Mmusculus.v79", quietly = TRUE))
BiocManager::install("EnsDb.Mmusculus.v79")
if (!requireNamespace("scater", quietly = TRUE))
BiocManager::install("scater")
if (!requireNamespace("bluster", quietly = TRUE))
BiocManager::install("bluster")
if (!requireNamespace("GenomeInfoDb", quietly = TRUE))
BiocManager::install("GenomeInfoDb")
if (!requireNamespace("GenomeInfoDb", quietly = TRUE))
BiocManager::install("GenomeInfoDb")
if (!requireNamespace("IRanges", quietly = TRUE))
BiocManager::install("IRanges")
if (!requireNamespace("ComplexHeatmap", quietly = TRUE))
BiocManager::install("ComplexHeatmap")
if (!requireNamespace("rtracklayer", quietly = TRUE))
BiocManager::install ("rtracklayer")
if (!requireNamespace("devtools", quietly = TRUE))
install.packages("devtools")
if (!requireNamespace("RColorBrewer", quietly = TRUE))
install.packages('RColorBrewer')
if (!requireNamespace("reticulate", quietly = TRUE))
install.packages("reticulate")
if (!requireNamespace("plyr", quietly = TRUE))
install.packages('plyr')
if (!requireNamespace("dsb", quietly = TRUE))
install.packages('dsb')
if (!requireNamespace("Seurat", quietly = TRUE))
install.packages('Seurat')
if (!requireNamespace("Signac", quietly = TRUE))
install.packages("Signac")
if (!requireNamespace("cluster", quietly = TRUE))
install.packages("cluster")
if (!requireNamespace("igraph", quietly = TRUE))
install.packages ("igraph")
if (!requireNamespace("ggplot2", quietly = TRUE))
install.packages ("ggplot2")
if (!requireNamespace("Matrix", quietly = TRUE))
install.packages ("Matrix")
if (!requireNamespace("dplyr", quietly = TRUE))
install.packages("dplyr")
if (!requireNamespace("tinytex", quietly = TRUE))
install.packages("tinytex")
if (!requireNamespace("tidyverse", quietly = TRUE))
install.packages ( "tidyverse" )
if (!requireNamespace("devtools", quietly = TRUE))
library(devtools)
if (!requireNamespace("MAESTRO", quietly = TRUE))
install_github("liulab-dfci/MAESTRO")
if (!requireNamespace("data.table", quietly = TRUE))
install.packages("data.table")
if (!requireNamespace("igraph", quietly = TRUE))
install.packages ("igraph")
if (!requireNamespace("parallel", quietly = TRUE))
install.packages("parallel")
if (!requireNamespace("dplyr", quietly = TRUE))
install.packages("dplyr")
if (!requireNamespace("Hmisc", quietly = TRUE))
install.packages("Hmisc")
if (!requireNamespace("CellChat", quietly = TRUE))
devtools::install_github("sqjin/CellChat")
if (!requireNamespace("patchwork", quietly = TRUE))
devtools::install_github("thomasp85/patchwork")
library(MAESTRO)
library(EnsDb.Hsapiens.v86)
library(EnsDb.Mmusculus.v79)
library(scater)
library(Seurat)
library(Signac)
library(cluster)
library(bluster)
library(GenomeInfoDb)
library(igraph)
library(GenomicRanges)
library(IRanges)
library(ggplot2)
library(Matrix)
library(dplyr)
library(tinytex)
library(tidyverse)
library(rtracklayer)
library(reticulate)
library(dplyr)
library(parallel)
library(igraph)
library(data.table)
library(Hmisc)
## ----------------------------------------------------------------------------------------------------------------
# Read data
##' Read matched scRNA + scATAC data from H5 file
#input:
# h5Path: the path of h5 file
# rna_matrix: an expression matrix with gene * cell
# atac_matrix: an accessibility matrix with peak * cell
# min_cell: the peak / gene will be removed if the value in the gene / peak with more than min_cell cell is equal to zero
# dataFormat: the format of input data, default as "h5"
#output:
#a seurat object
ReadData <- function(h5Path = NULL, rna_matrix = NULL, atac_matrix =NULL, min_cell = 0.1, dataFormat = "h5"){
if (dataFormat == "h5") {
tmp_obj <- Read10Xdata(h5Path = h5Path, min_cell = min_cell)
}else{
tmp_obj <- readmatrix(rna_matrix = rna_matrix, atac_matrix = atac_matrix, min_cell = min_cell)
}
tmp_obj <- filterCell(tmp_obj)
return(tmp_obj)
}
# Creat a seuart object
# input:
# h5Path: address of h5 file to read. (When dataFormat is 'h5', it couldn't be NULL.)
# rna_matrix: a RNA matrix where the rows correspond to genes and the columns correspond to cells. (When dataFormat is 'matrixs', it couldn't be NULL.)
# atac_matrix: a matrix where the rows correspond to peaks and the columns correspond to cells. (When data_type is 'RNA_ATAC'and dataFormat is 'matrixs', it couldn't be NULL.)
# adt_matrix: a matrix where the rows correspond to proteins and the columns correspond to cells. (When data_type is 'CITE' and dataFormat is 'matrixs', it couldn't be NULL.)
# data_type: 'CITE', 'RNA_ATAC'
# dataFormat: 'matrixs' or 'h5'
# min_cell: the peak/gene will be removed if the value in the gene/peak with more than min_cell cell is equal to zero
# nmad: a numeric scalar, specifying the minimum number of MADs away from median required for a value to be called an outlier
# gene.filter: if do gene filtering
# gene.filter: if do cell filtering
# output:
# obj: a seuart object for cite-seq data after normalizing and scaling (When data_type is 'CITE'); a seuart object for scRNA-seq and scATAC-seq (When data_type is 'RNA_ATAC')
ReadData <- function(h5Path = NULL, rna_matrix = NULL, atac_matrix = NULL, adt_matrix = NULL, data_type = NULL, dataFormat = NULL, min_cell=0.1, nmad=3, gene.filter=TRUE, cell.filter=TRUE){
if (data_type=='CITE'){
# obtain RNA and ADT matrixs
if (dataFormat=='matrixs'){
rna <- rna_matrix
adt <- adt_matrix
}else if(dataFormat=='h5'){
h5 <- Read10X_h5(h5Path)
rna <- h5$`Gene Expression`
adt <- h5$`Antibody Capture`
}
# gene filtering
if (gene.filter==TRUE){
binaryrna <- rna
binaryadt <- adt
binaryrna[binaryrna>0] <-1
binaryadt[binaryadt>0] <-1
rna <- rna[which(rowSums(binaryrna) > ncol(binaryrna)*min_cell),]
adt <- adt[which(rowSums(binaryadt) > ncol(binaryadt)*min_cell),]
}
#Setup a Seurat object, add the RNA and protein data
obj <- CreateSeuratObject(counts = rna)
obj [["percent.mt"]] <- PercentageFeatureSet(obj, pattern = "^MT-")
adt_assay <- CreateAssayObject(counts = adt)
obj[["ADT"]] <- adt_assay
# cell filtering
if (cell.filter==TRUE){
adtn<-isOutlier(
obj$nCount_ADT,
nmads = nmad,
log = F,
type = "both"
)
rnan<-isOutlier(
obj$nCount_RNA,
nmads = nmad,
log = F,
type = "both"
)
mito<-isOutlier(
obj$percent.mt,
nmads = nmad,
log = F,
type = "both"
)
obj <-
AddMetaData(obj, adtn, col.name = "adtn")
obj <-
AddMetaData(obj, rnan, col.name = "rnan")
obj <-
AddMetaData(obj, mito, col.name = "mito")
obj<-subset(
x = obj,
subset = adtn == F &
rnan == F &
mito == F
)
}
# normalization and scaling
DefaultAssay(obj) <- 'RNA'
obj <- NormalizeData(obj) %>% FindVariableFeatures()
all.genes <- rownames(obj)
obj <- ScaleData(obj, features = all.genes)
DefaultAssay(obj) <- 'ADT'
VariableFeatures(obj) <- rownames(obj[["ADT"]])
obj <- NormalizeData(obj, normalization.method = 'CLR', margin = 2) %>%
ScaleData()
DefaultAssay(obj) <- 'RNA'
}else if (data_type=='scRNA_scATAC'){
if (gene.filter==FALSE){
min_cell = 0
}
if (dataFormat == "h5") {
obj <- Read10Xdata(h5Path = h5Path, min_cell = min_cell)
}else{
obj <- readmatrix(rna_matrix = rna_matrix, atac_matrix = atac_matrix, min_cell = min_cell)
}
if (cell.filter==TRUE){
obj <- filterCell(obj, nmad= nmad)
}
}
return(obj)
}
# Read 10X data
##' Read matched scRNA + scATAC data from H5 file
#input:
# h5Path: the path of h5 file
# min_cell: the peak / gene will be removed if the value in the gene / peak with more than min_cell cell is equal to zero
#output:
# a seurat object
Read10Xdata <-
function(h5Path,
#annoObj = NULL,
#fragmentsPath = NULL,
#hintPath = NULL,
min_cell = 0.01) {
inputdata.10x <- Read10X_h5(h5Path)
rna_counts <- inputdata.10x$`Gene Expression`
atac_counts <- inputdata.10x$Peaks
grange.counts <-
StringToGRanges(rownames(atac_counts), sep = c(":", "-"))
grange.use <-
seqnames(grange.counts) %in% standardChromosomes(grange.counts)
atac_counts <- atac_counts[as.vector(grange.use), ]
chrom_assay <- CreateChromatinAssay(
counts = atac_counts,
sep = c(":", "-"),
min.cells = ncol(atac_counts) * min_cell,
#fragments = fragmentsPath,
#annotation = anno
#min.feature = 300,
)
tmp_obj <- CreateSeuratObject(counts = chrom_assay,
assay = "ATAC")
exp_assay <-
CreateAssayObject(counts = rna_counts,
min.cells = ncol(rna_counts) * min_cell)
tmp_obj[["RNA"]] <- exp_assay
DefaultAssay(tmp_obj) <- "RNA"
tmp_obj[["percent.mt"]] <-
PercentageFeatureSet(tmp_obj, pattern = "^MT-")
return (tmp_obj)
}
# Read data with matrix format
##' Read matched scRNA+scATAC data from matrix format
#input:
# rna_matrix: an expression matrix with gene * cell
# atac_matrix: an accessibility matrix with peak * cell
# min_cell: the peak/gene will be removed if the value in the gene/peak with more than min_cell cell is equal to zero
#output:
# a seurat object
readmatrix <- function(rna_matrix, atac_matrix, min_cell = 0.1) {
rna_matrix <-
rna_matrix[, intersect(colnames(atac_matrix), colnames(rna_matrix))]
atac_matrix <-
atac_matrix[, intersect(colnames(atac_matrix), colnames(rna_matrix))]
grange.counts <-
StringToGRanges(rownames(atac_matrix), sep = c(":", "-"))
grange.use <-
seqnames(grange.counts) %in% standardChromosomes(grange.counts)
atac_matrix <- atac_matrix[as.vector(grange.use), ]
atac_matrix <-
atac_matrix[lengths(strsplit(gsub(":", "-", rownames(atac_matrix)) , split = "-")) ==
3,]
rna_matrix <- rna_matrix[unique(rownames(rna_matrix)),]
#cell_type<-rna_cell$V7[-1][grepl("*RNA*",rna_gene$V2[-1])]
#min_cell=0.01
chrom_assay <- CreateChromatinAssay(
counts = atac_matrix,
sep = c(":", "-"),
#genome = annota,
#fragments = fragments,
min.cells = ncol(atac_matrix) * min_cell,
#min.feature = 300,
#annotation = annotations
)
obj <- CreateSeuratObject(counts = chrom_assay,
assay = "ATAC")
exp_assay <-
CreateAssayObject(counts = rna_matrix,
min.cells = ncol(rna_matrix) * min_cell)
obj[["RNA"]] <- exp_assay
DefaultAssay(obj) <- "RNA"
obj[["percent.mt"]] <-
PercentageFeatureSet(obj, pattern = "^mt-")
return(obj)
}
## ----------------------------------------------------------------------------------------------------------------
# Filter abnormal cells
##' Filter abnormal cells
#input:
# obj: a seurat object
# nmad: a numeric scalar, specifying the minimum number of MADs away from median required for a value to be called an outlier
# output:
# a seurat object
filterCell <- function(obj, nmad = 3, data_type = 'scRNA_scATAC') {
if (data_type == "scRNA_scATAC"){
atac <- isOutlier(obj$nCount_ATAC,
nmads = nmad,
log = F,
type = "both")
rna <- isOutlier(obj$nCount_RNA,
nmads = nmad,
log = F,
type = "both")
mito <- isOutlier(obj$percent.mt,
nmads = nmad,
log = F,
type = "both")
obj <-
AddMetaData(obj, atac, col.name = "atac")
obj <-
AddMetaData(obj, rna, col.name = "rna")
obj <-
AddMetaData(obj, mito, col.name = "mito")
obj <- subset(x = obj,
subset = atac == F &
rna == F &
mito == F)
}
if (data_type == "multipleRNA"){
rna<-isOutlier(
obj$nCount_RNA,
nmads = nmad,
log = F,
type = "both"
)
obj <-
AddMetaData(obj, rna, col.name = "rna")
obj<-subset(
x = obj,
subset = rna == F
)
}
return(obj)
}
## ----------------------------------------------------------------------------------------------------------------
##' Calculate gene active score matrix
# input:
# peak_count_matrix: a peak_count matrix from scATAC-seq with peak * cell which return from filterCell function
# organism: species type GRCh38 / GRCm38
# output:
# a gene * peak matrix, the elements represent the regulatory potential for peak to gene
CalGenePeakScore <-
function(peak_count_matrix, organism = "GRCh38") {
pbmc_peak <- peak_count_matrix
n <- nrow(pbmc_peak)
dia <- diag(n)
rownames(dia) <- rownames(pbmc_peak)
colnames(dia) <- 1:ncol(dia)
gene_peak <-
ATACCalculateGenescore(dia,
organism = organism,
decaydistance = 10000,
model = "Enhanced")
colnames(gene_peak) <- rownames(peak_count_matrix)
return (gene_peak)
}
## ----------------------------------------------------------------------------------------------------------------
##' Calculate gene active score matrix
#input:
# ATAC_gene_peak: a matrix with gene * peak which return from CalGenePeakScore fucntion
# obj: a seurat object after data preprocessing which return from filterCell function
# method: the method to integrate scRNA-seq and scATAC-seq velo (velocity) / WNN (weighted nearest neighbor)
# veloPath: if use velocity method, the veloPath should be provided
# return.weight: if return.weight = T, return modality integrated weight, else return GAS
#output:
# GAS matrix with gene * peak, the elements represent the gene activity score in each cell
# obj: a seurat object with obj[['ATAC_active']] a gene * peak matrix, the elements represent the regulatory potential for peak to gene
calculate_GAS_v1 <-
function(ATAC_gene_peak,
obj,
method = "velo",
return.weight = F,
veloPath = NULL) {
peak_count <- obj@assays$ATAC@counts
gene_count <- obj@assays$RNA@counts
peak_count[peak_count > 0] = 1
WA <- ATAC_gene_peak %*% peak_count
colnames(WA) <- colnames(peak_count)
rownames(WA) <- rownames(ATAC_gene_peak)
#print(colnames(WA)[1:3])
#print(colnames(obj)[1:3])
WA <- WA[which(rowSums(as.matrix(WA)) > 0),]
gene_count <-
gene_count[which(rowSums(as.matrix(gene_count)) > 0),]
commongene <-
intersect(x = rownames(WA), y = rownames(gene_count))
WA <- as.matrix(WA)
WA <- WA[commongene,]
atac_active <- CreateAssayObject(counts = WA,
min.cells = 0)
obj[['ATAC_active']] <- atac_active
gene_count <- gene_count[commongene,]
gene_rowsum <- rowSums(gene_count)
peak_rowsum <- rowSums(WA)
norm_gene_count <- gene_count / rowSums(gene_count)
norm_WBinary <- WA / rowSums(WA)
#norm_gene_count<-NormalizeData(CreateSeuratObject(counts = gene_count))$ RNA@data
gene_count <- norm_gene_count
#norm_WBinary<-NormalizeData(CreateSeuratObject(counts = WA))$RNA@data
peak_count <- norm_WBinary
#print(str(peak_count))
if (method == "velo") {
velo <- read.csv(veloPath, header = TRUE)
velo <- as.matrix(velo)
rownames(velo) <- velo[, 1]
temp <-
cbind(toupper(rownames(gene_count)), rownames(gene_count))
rownames(temp) <- temp[, 1]
temp <-
temp[intersect(toupper(rownames(gene_count)), rownames(velo)),]
velo <- velo[temp[, 1],]
rownames(velo) <- temp[, 2]
velo <- velo[,-1]
colnames(velo) <- gsub("\\.", "-", colnames(velo))
vv <- matrix(as.numeric(velo), dim(velo)[1], dim(velo)[2])
rownames(vv) <- rownames(velo)
colnames(vv) <- colnames(velo)
velo <- vv
rm(vv)
rna <- gene_count
atac <- peak_count
velo <-
velo[intersect(rownames(rna), rownames(velo)), intersect(colnames(rna), colnames(velo))]
rna <-
rna[intersect(rownames(rna), rownames(velo)), intersect(colnames(rna), colnames(velo))]
atac <-
atac[intersect(rownames(rna), rownames(velo)), intersect(colnames(rna), colnames(velo))]
gene_rowsum <- gene_rowsum[rownames(rna)]
peak_rowsum <- peak_rowsum[rownames(rna)]
#str(velo)
genes <- dim(velo)[1]
cells <- dim(velo)[2]
# rank matrix
rank_cell <- velo
rank_gene <- velo
rank_cell <- apply(velo, 2, rank)
rank_gene <- t(apply(velo, 1, rank))
rank_cell[velo > 0] = genes - rank_cell[velo > 0]
rank_cell[velo < 0] = rank_cell[velo < 0] - 1
rank_cell[velo == 0] = 0
rank_gene[velo > 0] = cells - rank_gene[velo > 0]
rank_gene[velo < 0] = rank_gene[velo < 0] - 1
rank_gene[velo == 0] = 0
# number of positive/negative for each gene/cell
cell_posi_num <- colSums(velo > 0)
cell_nega_num <- colSums(velo < 0)
gene_posi_num <- rowSums(velo > 0)
gene_nega_num <- rowSums(velo < 0)
# weights
weights = ((rank_cell ^ 2 + rank_gene ^ 2) / ((t((t(velo > 0)) * cell_posi_num + (t(velo <
0)) * cell_nega_num
)) ^ 2 + ((velo > 0) * gene_posi_num + (velo < 0) * gene_nega_num
) ^ 2 + (velo == 0))) ^ 0.5
weights[velo < 0] = weights[velo < 0] * (-1)
# GAS
GAS = rna * gene_rowsum + ((1 + weights) * atac) * ((1 + weights) *
peak_rowsum)
}
if (method == "wnn") {
obj <- obj[, colnames(gene_count)]
DefaultAssay(obj) <- "RNA"
obj <-
FindVariableFeatures(obj,
selection.method = "vst",
nfeatures = 2000)
obj <- ScaleData(obj, features = VariableFeatures(obj))
obj <- RunPCA(obj)
# ATAC analysis
# We exclude the first dimension as this is typically correlated with sequencing depth
DefaultAssay(obj) <- "ATAC"
obj <- RunTFIDF(obj)
obj <- FindTopFeatures(obj, min.cutoff = 'q0')
obj <- RunSVD(obj)
obj <-
RunUMAP(
obj,
reduction = 'lsi',
dims = 2:50,
reduction.name = "umap.atac",
reduction.key = "atacUMAP_"
)
obj <-
FindMultiModalNeighbors(obj,
reduction.list = list("pca", "lsi"),
dims.list = list(1:50, 2:50))
GAS <-
gene_count * gene_rowsum * obj$RNA.weight + peak_count * obj$ATAC.weight *
peak_rowsum
}
if (isTRUE(return.weight)) {
return(weights)
} else {
m <- list()
m[[1]] <- GAS
m[[2]] <- obj
return(m)
}
}
# required package -- reticulate
# input:
# GAS: the spliced and normalized matrix obtained from CLR function
# result_dir: The address for storing the models and optimization results(Type:str)
# epoch:(Type:int)
# lr: learning rate(Type:float)
# n_hid: Number of hidden dimension(Type:int)
# n_heads: Number of attention head(Type:int)
# cuda: 0 use GPU0 else cpu(Type:int)
# data_type: 'CITE', 'RNA_ATAC', or 'multiple RNA'
# envPath: The address for environment to use if use.env is TRUE(Type:str)
# output:
# HGT_result: a list containing requried results of HGT model as follows:
# parameters: given parameters from user --epoch, lr, n_hid, n_heads, cuda
# cell_hgt_matrix: cell embedding matrix
# feature_hgt_matrix : gene embedding matrix and protein embedding matrix when data_type is 'CITE';
# attention: attention meassage for features and cells
# data_type: 'CITE', 'RNA_ATAC', or 'multiple RNA'
# result_dir: The address for storing the models and optimization results
# GAS: the spliced and normalized matrix obtained from CLR function
run_HGT <- function(GAS,result_dir,data_type,envPath=NULL,lr=NULL, epoch=NULL, n_hid=NULL, n_heads=NULL,cuda=0){
if (data_type == 'CITE') {
if (is.null(lr)){lr = 0.1}
if (is.null(epoch)){epoch = 50}
if (is.null(n_hid)){n_hid = 104}
if (is.null(n_heads)){n_heads = 13}
}
if (data_type == 'scRNA_scATAC') {
if (is.null(lr)){lr = 0.1}
if (is.null(epoch)){epoch = 100}
if (is.null(n_hid)){n_hid = 128}
if (is.null(n_heads)){n_heads = 16}
}
if (data_type == 'multipleRNA') {
if (is.null(lr)){lr = 0.1}
if (is.null(epoch)){epoch = 100}
if (is.null(n_hid)){n_hid = 104}
if (is.null(n_heads)){n_heads = 13}
}
print(epoch)
cat(lr, epoch, n_hid, n_heads, cuda)
if (!is.null(envPath)){use_condaenv(envPath)}
list_in <- assign("list_in", list(lr=lr, epoch=epoch, n_hid=n_hid, n_heads=n_heads, result_dir=result_dir, cuda=cuda, data_type=data_type, cell_gene=GAS, gene_name=rownames(GAS), cell_name=colnames(GAS)), envir = .GlobalEnv)
source_python('./arg.py')
cell_hgt_matrix <- py$cell_matrix
gene_hgt_matrix <- py$gene_matrix
attention <- py$df2
rownames(cell_hgt_matrix) <- list_in$cell_name
rownames(gene_hgt_matrix) <- list_in$gene_name
HGT_result <- list()
HGT_result[['parameters']] <- data.frame(lr,epoch,n_hid,n_heads,cuda)
HGT_result[['GAS']] <- GAS
HGT_result[['cell_hgt_matrix']] <- cell_hgt_matrix
HGT_result[['feature_hgt_matrix']] <- gene_hgt_matrix
HGT_result[['attention']] <- attention
HGT_result[['result_dir']] <- result_dir
HGT_result[['data_type']] <- data_type
return(HGT_result)
}
## ----------------------------------------------------------------------------------------------------------------
# CT active gene modules calculation
# input:
# GAS: a gene active matrix with gene * cell which return from calculate_GAS function
# cell_hgt_matrix: cell-embedding matrix which otains from HGT function
# att: attention matrix with gene-cell * head which obtain from HGT function
# gene_hgt_matrix: gene_embedding matrix with gene-cell * head which obtain from HGT function
# cutoff: the threshold of gene module scale, the higher value the fewer number gene in the module. Default as 1.6.
#output:
# co (variable 1): a biological gene module. a list with name CT-i and active gene list in CT-i
get_gene_module <-
function(obj, GAS, att, cutoff = 1.6, method = NULL) {
if (method == 'SFp'){
`%!in%` <- Negate(`%in%`) # define the negation of %in%
n.matching <- 10 # the number of predicting maximum matching
m<-inp(GAS, HGT_result[['attention']], HGT_result[['cell_hgt_matrix']], HGT_result[['feature_hgt_matrix']], l=1.2)
df1 <- m[[1]]
df2 <- m[[2]]
terminals <- m[[3]]
cells <- m[[4]]
graph.out<-m[[5]]
cut.G <- global_matching_graph(df1, df2, n.matching = 10, cells = cells, terminals)
cat ('The cutted graph contains', length(V(cut.G)), 'nodes and', length(E(cut.G)), 'edges.\n')
steiner.ig <- set_cover_mst(G = cut.G, terminals = terminals)
mods <- get_modules(steiner.ig = steiner.ig, terminals = terminals,
out.file = out.file)
co<-list()
for (i in unique(mods$terminal)){
a<-as.numeric(unlist(strsplit(as.character(unique(mods$terminal)[i]), split ='[.]'))[1])-1
co[[paste0('ct_',a)]]<-unique(mods$steiner_node[mods$terminal==i])
}
}else{
graph.out <- Idents(obj)
nhead <- ncol(att)
gene_name <- rownames(GAS)[att$gene + 1]
cell_name <- colnames(GAS)[att$cell + 1]
att$ct <- graph.out[cell_name]
att$gene_name <- gene_name
att$cell_name <- cell_name
mod <- function(x) {
return(sqrt(sum(c(x ^ 2))))
}
nor <- function(x) {
return((x - min(x)) / (max(x) - min(x)))
}
att[, 3:nhead] <- nor(att[, 3:nhead])
attention <-
aggregate(x = as.list(att[, 3:nhead]),
by = list(att$ct, att$gene_name),
mean)
#att[,4:nhead]<-1-att[,4:nhead]
weight <- apply(att[, 3:nhead], 1, mod)
df <-
data.frame(
'node1' = att$gene_name,
'node2' = att$cell_name,
'weight' = weight,
'ct' = att$ct
)
attention <-
aggregate(x = df$weight, by = list(df$ct, df$node1), mean)
co <- list()
for (i in (0:(length(unique(att$ct)) - 1))) {
t <-
mean(attention[attention$Group.1 == i, ]$x) + 1.6 * sd(attention[attention$Group.1 ==
i, ]$x)
co[[paste('ct', i, sep = "_")]] <-
attention[attention$Group.1 == i, ]$Group.2[attention[attention$Group.1 ==
i, ]$x > t]
}
}
return (co)
}
## ----------------------------------------------------------------------------------------------------------------
# gene module save
#input:
# co: the active gene module from get_gene_module function
# lisa_path: the path of active gene module to save
#result
# write gene module to the lisa_path
write_GM <- function(co, lisa_path) {
if (length(dir(path = lisa_path, pattern = ".csv")) >
0) {
system(paste0("rm ", lisa_path, "*.csv "))
}
if (length(dir(path = lisa_path, pattern = ".txt")) >
0) {
system(paste0("rm ", lisa_path, "*.txt "))
}
for (j in (1:length(co))) {
if (length(unique(co[[j]])) < 20 |
length(unique(co[[j]])) > 20000) {
next
} else{
ct <- unlist(strsplit(names(co[j]), split = "_"))[1]
write.table(
co[[j]],
paste0(lisa_path, names(co[j]), ".txt"),
quote = F,
sep = "\t",
row.names = F,
col.names = F
)
}
}
}
## ----------------------------------------------------------------------------------------------------------------
# Filter gene with no accessible peak in promoter
# input:
# obj: a seurat object which return from filterCell function
# gene_peak: a matrix with gene * peak from scATAC-seq which return from filterCell function
# GAS: the GAS matrix with gene * cell which return calculate_GAS function
# species: human / mouse
#output:
# a matrix with gene * peak. The gene with no accessible peak will be removed
AccPromoter <- function(obj, gene_peak, GAS, species = "hg38") {
peak_cell <- obj@assays$ATAC@counts
if (species == "hg38") {
gene.ranges <- genes(EnsDb.Hsapiens.v86)
} else{
gene.ranges <- genes(EnsDb.Mmusculus.v79)
}
gene.use <-
seqnames(gene.ranges) %in% standardChromosomes(gene.ranges)[standardChromosomes(gene.ranges) !=
"MT"]
gene.ranges <- gene.ranges[as.vector(gene.use)]
gene.ranges <-
gene.ranges[gene.ranges$gene_name %in% rownames(GAS)]
genebodyandpromoter.coords <-
Extend(x = gene.ranges,
upstream = 2000,
downstream = 0)
#str(genebodyandpromoter.coords)
x <- as.data.frame(genebodyandpromoter.coords@ranges)
peaks <-
GRanges(
seqnames = paste("chr", genebodyandpromoter.coords@seqnames, sep = ""),
ranges = IRanges(start = , x$start,
width = x$width)
)
peak_name <-
colnames(gene_peak)[lengths(strsplit(gsub(":", "-", colnames(gene_peak)) , split = "-")) ==
3]
peak_name <-
do.call(what = rbind, strsplit(gsub(":", "-", peak_name) , split = "-"))
peak_name <- as.data.frame(peak_name)
names(peak_name) <- c("chromosome", 'start', 'end')
peak_name <- GenomicRanges::makeGRangesFromDataFrame(peak_name)
#str(peaks)
over <- findOverlaps(peak_name, peaks)
#str(over)
promoter_gene <-
genebodyandpromoter.coords$gene_name[unique(over@to)]
str(promoter_gene)
gene_peak <- gene_peak[promoter_gene,]
return(gene_peak)
}
## ----------------------------------------------------------------------------------------------------------------
##' infer ct active regulons
# input:
# GAS: the GAS matrix with gene * cell which return calculate_GAS function
# co: a list of bio network which reture from gene_ function
# gene_peak_pro: the matrix with gene * peak which return AccPromoter function
# species: human / mouse (human = "hg38", mouse = "mm10" )
# humanPath: if species == human, the TF binding RData absolute path of hg38 should be provided
# mousePath: if species == mouse, the TF binding RData absolute path of mm10 should be provided
#output:
# BA_score a TF binding affinity matrix with TF * peak, the elements in the matrix is the binding power of TF to peak
# ct_regulon: candidate cell type active regulon
# TFinGAS: TRUE / FALSE, if number of the intersection of candidate TF from LISA and gene in GAS > 50 TFinGAS will be true, else it will be false
Calregulon <-
function(GAS,
co,
gene_peak_pro,
species = "hg38",
jaspar_path = "/scratch/deepmaps/jaspar",
lisa_path = "/home/wan268/hgt/RNA_ATAC/lymph_14k/") {
if (species == "hg38") {
tfbs_df <- qs::qread(paste0(jaspar_path, "jaspar_hg38_500.qsave"))
}
else {
tfbs_df <- qs::qread(paste0(jaspar_path, "jaspar_mm10_500.qsave"))
}
BA_score <-
matrix(0, ncol(gene_peak_pro), length(unique(tfbs_df$V4)))
colnames(BA_score) <- unique(tfbs_df$V4)
rownames(BA_score) <- colnames(gene_peak_pro)
gene_TF <-
matrix(0, nrow(gene_peak_pro), length(unique(tfbs_df$V4)))
colnames(gene_TF) <- unique(tfbs_df$V4)
rownames(gene_TF) <- rownames(gene_peak_pro)
peak <- tfbs_df[, 1:3]
colnames(peak) <- c("chromosome", 'start', 'end')
peak <- GenomicRanges::makeGRangesFromDataFrame(peak)
ct_subregulon <- list()
ct_regulon <- list()
coexp_tf <- list()
No <- 1
for (i in (1:length(co))) {
if (length(co[[i]]) > 0) {
co[[i]] <- intersect(co[[i]], rownames(gene_peak_pro))
a <- which(gene_peak_pro[co[[i]],] > 0, arr.ind = T)
op <- colnames(gene_peak_pro)[unname(a[, 'col'])]
peak_name <-
op[lengths(strsplit(gsub(":", "-", op) , split = "-")) == 3]
peak_name <-
do.call(what = rbind, strsplit(gsub(":", "-", peak_name) , split = "-"))
peak_name <- as.data.frame(peak_name)
names(peak_name) <- c("chromosome", 'start', 'end')
peak_name <-
GenomicRanges::makeGRangesFromDataFrame(peak_name)
over <- findOverlaps(peak_name, peak)
#print(i)
p <- op[over@from]
pp <- tfbs_df$V5[over@to] / 100
df <- data.frame(p, pp, tfbs_df$V4[over@to])
hh <- df[!duplicated(df[,-2]),]
for (k1 in (1:nrow(hh))) {
BA_score[hh[k1,]$p, hh[k1,]$tfbs_df.V4.over.to.] <- hh[k1,]$pp
}
gene_TF <- gene_peak_pro %*% BA_score
TF <- unique(tfbs_df$V4[over@to])
if (length(co[[i]]) < 20000 & length(co[[i]]) > 20) {
tf <-
read.csv(paste(lisa_path,
names(co[i]),
".txt.csv",
sep = ""))
tf_pval_0.05 <- unique(tf[, 3][tf$summary_p_value < 0.05])
TF <- intersect(unique(TF), tf_pval_0.05)
}
#print(length(TF))
#gene_TF[co[[i]],pp]<-gene_peak_pro[co[[i]],p] %*% BA_score[p,TF]
coexp_tf[[names(co[i])]] <- TF
#print(length(TF))
if (length(TF) > 0) {
for (k in 1:length(TF)) {
if (length(intersect(TF[k], rownames(GAS))) > 50) {
TFinGAS <- T
if (TF[k] %in% rownames(GAS)) {
#print(TF[k])
a <- unlist(strsplit(names(co[i]), "_"))
a <- paste0(a[1], a[2])
h <- paste(TF[k], a, sep = "_")
ct_subregulon[[h]] <-
co[[i]][gene_TF[co[[i]], TF[k]] > 0]
}
} else{
a <- unlist(strsplit(names(co[i]), "_"))
TFinGAS <- F
a <- paste0(a[1], a[2])
h <- paste(TF[k], a, sep = "_")
ct_subregulon[[h]] <-
co[[i]][gene_TF[co[[i]], TF[k]] > 0]
}
}
}
}
}
m <- list()
m[[1]] <- BA_score
ct_subregulon <- ct_subregulon[lengths(ct_subregulon) > 10]
m[[2]] <- ct_subregulon
m[[3]] <- TFinGAS
return(m)
}
## ----------------------------------------------------------------------------------------------------------------
# combine same TF
#input:
# gene_peak_pro: a matrix with gene * peak. The gene with no accessible peak will be removed which return from AccPromoter function
# BA_score: a TF binding affinity matrix with TF * peak, the elements in the matrix is the binding power of TF to peak which returen from Calregulon function
#output:
# peak_TF: a matrix with peak * TF without repeat TF
uni <- function(gene_peak_pro, BA_score) {
gene_TF <- gene_peak_pro %*% BA_score
rownames(gene_TF) <- rownames(gene_peak_pro)
mat <-
matrix(0, nrow = length(rownames(gene_TF)), ncol = length(unique(colnames(gene_TF))))#peak_TF
mat1 <-
matrix(0, nrow = length(rownames(BA_score)), ncol = length(unique(colnames(BA_score))))#gene_TF
rownames(mat1) <- rownames(BA_score)
colnames(mat1) <- unique(colnames(BA_score))
#mat1: peak_TF score
for (x in unique(colnames(BA_score))) {
if (is.null(nrow(gene_TF[, colnames(BA_score) == x]))) {
#mat<-rbind(mat, unlist(gene_peak_matrix[rownames(gene_peak_matrix)==x,]))
mat1[, x] <-
unname(unlist(BA_score[, colnames(BA_score) == x]))
} else{
#mat<-rbind(mat,unlist(colSums(gene_peak_matrix[rownames(gene_peak_matrix)==x,])))
mat1[, x] <-
unname(unlist(rowSums(BA_score[, colnames(BA_score) == x])))
}
}
rownames(mat) <- rownames(gene_TF)
colnames(mat) <- unique(colnames(gene_TF))
for (x in unique(colnames(gene_TF))) {
if (is.null(nrow(gene_TF[, colnames(gene_TF) == x]))) {
#print("111")
#mat<-rbind(mat, unlist(gene_peak_matrix[rownames(gene_peak_matrix)==x,]))
mat[, x] <- unname(unlist(gene_TF[, colnames(gene_TF) == x]))
} else{
#mat<-rbind(mat,unlist(colSums(gene_peak_matrix[rownames(gene_peak_matrix)==x,])))
mat[, x] <-
unname(unlist(rowSums(gene_TF[, colnames(gene_TF) == x])))
}
}
#m = list()
#m[[1]] <- mat
#m[[2]] <- mat1
return (mat1)
}
## ----------------------------------------------------------------------------------------------------------------
# Regulatory Intensive (RI) score in cell level
# input:
# 1 - obj: a seurat object which return from filterCell function
# 2 - ct_regulon: cell type active regulon
# 3 - GAS: the GAS matrix with gene * cell which return calculate_GAS function
# 4 - gene_peak_pro: the matrix with gene * peak which return from AccPromoter function
# 5 - peak_TF: the matrix with peak * TF which return from uni function
# 6 - graph.out: a factor variable. The predict cell cluster which return from get_gene_module function
#output:
# - RI_C: a regulatory intensive matrix with TF-gene pair * cell, the element means the intensity of TF to gene in each cell
RI_cell <-
function(obj,
ct_regulon,
GAS,
gene_peak_pro,
peak_TF,
graph.out) {
v <- vector()
for (i in (1:length(ct_regulon))) {
v <-
append(v, paste(unlist(strsplit(
names(ct_regulon[i]), "_"
))[1], ct_regulon[[i]], sep = "_"))
}
peak_cell <- obj@assays$ATAC@counts
peak_cell[peak_cell > 0] = 1
peak_cell <- (peak_cell[, colnames(GAS)])
#graph.out<-Idents(obj)
#TG_cell <- foreach(i=1:length(ct_regulon), .packages='Matrix',.combine='c') %dopar%{
TG_cell <- matrix(0, length(unique(v)), length(graph.out))