forked from satijalab/seurat
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathpreprocessing.R
2919 lines (2871 loc) · 99.5 KB
/
preprocessing.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
#' @include generics.R
#'
NULL
#%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
# Functions
#%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
#' Calculate the Barcode Distribution Inflection
#'
#' This function calculates an adaptive inflection point ("knee") of the barcode distribution
#' for each sample group. This is useful for determining a threshold for removing
#' low-quality samples.
#'
#' The function operates by calculating the slope of the barcode number vs. rank
#' distribution, and then finding the point at which the distribution changes most
#' steeply (the "knee"). Of note, this calculation often must be restricted as to the
#' range at which it performs, so `threshold` parameters are provided to restrict the
#' range of the calculation based on the rank of the barcodes. [BarcodeInflectionsPlot()]
#' is provided as a convenience function to visualize and test different thresholds and
#' thus provide more sensical end results.
#'
#' See [BarcodeInflectionsPlot()] to visualize the calculated inflection points and
#' [SubsetByBarcodeInflections()] to subsequently subset the Seurat object.
#'
#' @param object Seurat object
#' @param barcode.column Column to use as proxy for barcodes ("nCount_RNA" by default)
#' @param group.column Column to group by ("orig.ident" by default)
#' @param threshold.high Ignore barcodes of rank above thisf threshold in inflection calculation
#' @param threshold.low Ignore barcodes of rank below this threshold in inflection calculation
#'
#' @return Returns Seurat object with a new list in the `tools` slot, `CalculateBarcodeInflections` with values:
#'
#' * `barcode_distribution` - contains the full barcode distribution across the entire dataset
#' * `inflection_points` - the calculated inflection points within the thresholds
#' * `threshold_values` - the provided (or default) threshold values to search within for inflections
#' * `cells_pass` - the cells that pass the inflection point calculation
#'
#' @importFrom methods slot
#' @importFrom stats ave aggregate
#'
#' @export
#'
#' @author Robert A. Amezquita, \email{[email protected]}
#' @seealso \code{\link{BarcodeInflectionsPlot}} \code{\link{SubsetByBarcodeInflections}}
#'
#' @examples
#' CalculateBarcodeInflections(pbmc_small, group.column = 'groups')
#'
CalculateBarcodeInflections <- function(
object,
barcode.column = "nCount_RNA",
group.column = "orig.ident",
threshold.low = NULL,
threshold.high = NULL
) {
## Check that barcode.column exists in meta.data
if (!(barcode.column %in% colnames(x = object[[]]))) {
stop("`barcode.column` specified not present in Seurat object provided")
}
# Calculation of barcode distribution
## Append rank by grouping x umi column
# barcode_dist <- as.data.frame([email protected])[, c(group.column, barcode.column)]
barcode_dist <- object[[c(group.column, barcode.column)]]
barcode_dist <- barcode_dist[do.call(what = order, args = barcode_dist), ] # order by columns left to right
barcode_dist$rank <- ave(
x = barcode_dist[, barcode.column], barcode_dist[, group.column],
FUN = function(x) {
return(rev(x = order(x)))
}
)
barcode_dist <- barcode_dist[order(barcode_dist[, group.column], barcode_dist[, 'rank']), ]
## calculate rawdiff and append per group
top <- aggregate(
x = barcode_dist[, barcode.column],
by = list(barcode_dist[, group.column]),
FUN = function(x) {
return(c(0, diff(x = log10(x = x + 1))))
})$x
bot <- aggregate(
x = barcode_dist[, 'rank'],
by = list(barcode_dist[, group.column]),
FUN = function(x) {
return(c(0, diff(x = x)))
}
)$x
barcode_dist$rawdiff <- unlist(x = mapply(
FUN = function(x, y) {
return(ifelse(test = is.na(x = x / y), yes = 0, no = x / y))
},
x = top,
y = bot
))
# Calculation of inflection points
## Set thresholds for rank of barcodes to ignore
threshold.low <- threshold.low %||% 1
threshold.high <- threshold.high %||% max(barcode_dist$rank)
## Subset the barcode distribution by thresholds
barcode_dist_sub <- barcode_dist[barcode_dist$rank > threshold.low & barcode_dist$rank < threshold.high, ]
## Calculate inflection points
## note: if thresholds are s.t. it produces the same length across both groups,
## aggregate will create a data.frame with x.* columns, where * is the length
## using the same combine approach will yield non-symmetrical results!
whichmin_list <- aggregate(
x = barcode_dist_sub[, 'rawdiff'],
by = list(barcode_dist_sub[, group.column]),
FUN = function(x) {
return(x == min(x))
}
)$x
## workaround for aggregate behavior noted above
if (is.list(x = whichmin_list)) { # uneven lengths
is_inflection <- unlist(x = whichmin_list)
} else if (is.matrix(x = whichmin_list)) { # even lengths
is_inflection <- as.vector(x = t(x = whichmin_list))
}
tmp <- cbind(barcode_dist_sub, is_inflection)
# inflections <- tmp[tmp$is_inflection == TRUE, c(group.column, barcode.column, "rank")]
inflections <- tmp[which(x = tmp$is_inflection), c(group.column, barcode.column, 'rank')]
# Use inflection point for what cells to keep
## use the inflection points to cut the subsetted dist to what to keep
## keep only the barcodes above the inflection points
keep <- unlist(x = lapply(
X = whichmin_list,
FUN = function(x) {
keep <- !x
if (sum(keep) == length(x = keep)) {
return(keep) # prevents bug in case of keeping all cells
}
# toss <- which(keep == FALSE):length(x = keep) # the end cells below knee
toss <- which(x = !keep):length(x = keep)
keep[toss] <- FALSE
return(keep)
}
))
barcode_dist_sub_keep <- barcode_dist_sub[keep, ]
cells_keep <- rownames(x = barcode_dist_sub_keep)
# Bind thresholds to keep track of where they are placed
thresholds <- data.frame(
threshold = c('threshold.low', 'threshold.high'),
rank = c(threshold.low, threshold.high)
)
# Combine relevant info together
## Combine Barcode dist, inflection point, and cells to keep into list
info <- list(
barcode_distribution = barcode_dist,
inflection_points = inflections,
threshold_values = thresholds,
cells_pass = cells_keep
)
# save results into object
Tool(object = object) <- info
return(object)
}
#' Convert a peak matrix to a gene activity matrix
#'
#' This function will take in a peak matrix and an annotation file (gtf) and collapse the peak
#' matrix to a gene activity matrix. It makes the simplifying assumption that all counts in the gene
#' body plus X kb up and or downstream should be attributed to that gene.
#'
#' @param peak.matrix Matrix of peak counts
#' @param annotation.file Path to GTF annotation file
#' @param seq.levels Which seqlevels to keep (corresponds to chromosomes usually)
#' @param include.body Include the gene body?
#' @param upstream Number of bases upstream to consider
#' @param downstream Number of bases downstream to consider
#' @param verbose Print progress/messages
#'
#' @importFrom future nbrOfWorkers
#' @export
#'
CreateGeneActivityMatrix <- function(
peak.matrix,
annotation.file,
seq.levels = c(1:22, "X", "Y"),
include.body = TRUE,
upstream = 2000,
downstream = 0,
verbose = TRUE
) {
if (!PackageCheck('GenomicRanges', error = FALSE)) {
stop("Please install GenomicRanges from Bioconductor.")
}
if (!PackageCheck('rtracklayer', error = FALSE)) {
stop("Please install rtracklayer from Bioconductor.")
}
# convert peak matrix to GRanges object
peak.df <- rownames(x = peak.matrix)
peak.df <- do.call(what = rbind, args = strsplit(x = gsub(peak.df, pattern = ":", replacement = "-"), split = "-"))
peak.df <- as.data.frame(x = peak.df)
colnames(x = peak.df) <- c("chromosome", 'start', 'end')
peaks.gr <- GenomicRanges::makeGRangesFromDataFrame(df = peak.df)
# if any peaks start at 0, change to 1
# otherwise GenomicRanges::distanceToNearest will not work
BiocGenerics::start(peaks.gr[BiocGenerics::start(peaks.gr) == 0, ]) <- 1
# get annotation file, select genes
gtf <- rtracklayer::import(con = annotation.file)
gtf <- GenomeInfoDb::keepSeqlevels(x = gtf, value = seq.levels, pruning.mode = 'coarse')
# change seqlevelsStyle if not the same
if (!any(GenomeInfoDb::seqlevelsStyle(x = gtf) == GenomeInfoDb::seqlevelsStyle(x = peaks.gr))) {
GenomeInfoDb::seqlevelsStyle(gtf) <- GenomeInfoDb::seqlevelsStyle(peaks.gr)
}
gtf.genes <- gtf[gtf$type == 'gene']
# Extend definition up/downstream
if (include.body) {
gtf.body_prom <- Extend(x = gtf.genes, upstream = upstream, downstream = downstream)
} else {
gtf.body_prom <- SummarizedExperiment::promoters(x = gtf.genes, upstream = upstream, downstream = downstream)
}
gene.distances <- GenomicRanges::distanceToNearest(x = peaks.gr, subject = gtf.body_prom)
keep.overlaps <- gene.distances[rtracklayer::mcols(x = gene.distances)$distance == 0]
peak.ids <- peaks.gr[S4Vectors::queryHits(x = keep.overlaps)]
gene.ids <- gtf.genes[S4Vectors::subjectHits(x = keep.overlaps)]
# Some GTF rows will not have gene_name attribute
# Replace it by gene_id attribute
gene.ids$gene_name[is.na(gene.ids$gene_name)] <- gene.ids$gene_id[is.na(gene.ids$gene_name)]
peak.ids$gene.name <- gene.ids$gene_name
peak.ids <- as.data.frame(x = peak.ids)
peak.ids$peak <- rownames(peak.matrix)[S4Vectors::queryHits(x = keep.overlaps)]
annotations <- peak.ids[, c('peak', 'gene.name')]
colnames(x = annotations) <- c('feature', 'new_feature')
# collapse into expression matrix
peak.matrix <- as(object = peak.matrix, Class = 'matrix')
all.features <- unique(x = annotations$new_feature)
if (nbrOfWorkers() > 1) {
mysapply <- future_sapply
} else {
mysapply <- ifelse(test = verbose, yes = pbsapply, no = sapply)
}
newmat <- mysapply(X = 1:length(x = all.features), FUN = function(x){
features.use <- annotations[annotations$new_feature == all.features[[x]], ]$feature
submat <- peak.matrix[features.use, ]
if (length(x = features.use) > 1) {
return(Matrix::colSums(x = submat))
} else {
return(submat)
}
})
newmat <- t(x = newmat)
rownames(x = newmat) <- all.features
colnames(x = newmat) <- colnames(x = peak.matrix)
return(as(object = newmat, Class = 'dgCMatrix'))
}
#' Demultiplex samples based on data from cell 'hashing'
#'
#' Assign sample-of-origin for each cell, annotate doublets.
#'
#' @param object Seurat object. Assumes that the hash tag oligo (HTO) data has been added and normalized.
#' @param assay Name of the Hashtag assay (HTO by default)
#' @param positive.quantile The quantile of inferred 'negative' distribution for each hashtag - over which the cell is considered 'positive'. Default is 0.99
#' @param init Initial number of clusters for hashtags. Default is the # of hashtag oligo names + 1 (to account for negatives)
#' @param kfunc Clustering function for initial hashtag grouping. Default is "clara" for fast k-medoids clustering on large applications, also support "kmeans" for kmeans clustering
#' @param nsamples Number of samples to be drawn from the dataset used for clustering, for kfunc = "clara"
#' @param nstarts nstarts value for k-means clustering (for kfunc = "kmeans"). 100 by default
#' @param seed Sets the random seed. If NULL, seed is not set
#' @param verbose Prints the output
#'
#' @return The Seurat object with the following demultiplexed information stored in the meta data:
#' \describe{
#' \item{hash.maxID}{Name of hashtag with the highest signal}
#' \item{hash.secondID}{Name of hashtag with the second highest signal}
#' \item{hash.margin}{The difference between signals for hash.maxID and hash.secondID}
#' \item{classification}{Classification result, with doublets/multiplets named by the top two highest hashtags}
#' \item{classification.global}{Global classification result (singlet, doublet or negative)}
#' \item{hash.ID}{Classification result where doublet IDs are collapsed}
#' }
#'
#' @importFrom cluster clara
#' @importFrom Matrix colSums
#' @importFrom fitdistrplus fitdist
#' @importFrom stats pnbinom kmeans
#'
#' @export
#'
#' @seealso \code{\link{HTOHeatmap}}
#'
#' @examples
#' \dontrun{
#' object <- HTODemux(object)
#' }
#'
HTODemux <- function(
object,
assay = "HTO",
positive.quantile = 0.99,
init = NULL,
nstarts = 100,
kfunc = "clara",
nsamples = 100,
seed = 42,
verbose = TRUE
) {
if (!is.null(x = seed)) {
set.seed(seed = seed)
}
#initial clustering
assay <- assay %||% DefaultAssay(object = object)
data <- GetAssayData(object = object, assay = assay)
counts <- GetAssayData(
object = object,
assay = assay,
slot = 'counts'
)[, colnames(x = object)]
counts <- as.matrix(x = counts)
ncenters <- init %||% (nrow(x = data) + 1)
switch(
EXPR = kfunc,
'kmeans' = {
init.clusters <- kmeans(
x = t(x = GetAssayData(object = object, assay = assay)),
centers = ncenters,
nstart = nstarts
)
#identify positive and negative signals for all HTO
Idents(object = object, cells = names(x = init.clusters$cluster)) <- init.clusters$cluster
},
'clara' = {
#use fast k-medoid clustering
init.clusters <- clara(
x = t(x = GetAssayData(object = object, assay = assay)),
k = ncenters,
samples = nsamples
)
#identify positive and negative signals for all HTO
Idents(object = object, cells = names(x = init.clusters$clustering), drop = TRUE) <- init.clusters$clustering
},
stop("Unknown k-means function ", kfunc, ", please choose from 'kmeans' or 'clara'")
)
#average hto signals per cluster
#work around so we don't average all the RNA levels which takes time
average.expression <- AverageExpression(
object = object,
assays = assay,
verbose = FALSE
)[[assay]]
#checking for any cluster with all zero counts for any barcode
if (sum(average.expression == 0) > 0) {
stop("Cells with zero counts exist as a cluster.")
}
#create a matrix to store classification result
discrete <- GetAssayData(object = object, assay = assay)
discrete[discrete > 0] <- 0
# for each HTO, we will use the minimum cluster for fitting
for (iter in rownames(x = data)) {
values <- counts[iter, colnames(object)]
#commented out if we take all but the top cluster as background
#values_negative=values[setdiff([email protected],WhichCells(object,which.max(average.expression[iter,])))]
values.use <- values[WhichCells(
object = object,
idents = levels(x = Idents(object = object))[[which.min(x = average.expression[iter, ])]]
)]
fit <- suppressWarnings(expr = fitdist(data = values.use, distr = "nbinom"))
cutoff <- as.numeric(x = quantile(x = fit, probs = positive.quantile)$quantiles[1])
discrete[iter, names(x = which(x = values > cutoff))] <- 1
if (verbose) {
message(paste0("Cutoff for ", iter, " : ", cutoff, " reads"))
}
}
# now assign cells to HTO based on discretized values
npositive <- colSums(x = discrete)
classification.global <- npositive
classification.global[npositive == 0] <- "Negative"
classification.global[npositive == 1] <- "Singlet"
classification.global[npositive > 1] <- "Doublet"
donor.id = rownames(x = data)
hash.max <- apply(X = data, MARGIN = 2, FUN = max)
hash.maxID <- apply(X = data, MARGIN = 2, FUN = which.max)
hash.second <- apply(X = data, MARGIN = 2, FUN = MaxN, N = 2)
hash.maxID <- as.character(x = donor.id[sapply(
X = 1:ncol(x = data),
FUN = function(x) {
return(which(x = data[, x] == hash.max[x])[1])
}
)])
hash.secondID <- as.character(x = donor.id[sapply(
X = 1:ncol(x = data),
FUN = function(x) {
return(which(x = data[, x] == hash.second[x])[1])
}
)])
hash.margin <- hash.max - hash.second
doublet_id <- sapply(
X = 1:length(x = hash.maxID),
FUN = function(x) {
return(paste(sort(x = c(hash.maxID[x], hash.secondID[x])), collapse = "_"))
}
)
# doublet_names <- names(x = table(doublet_id))[-1] # Not used
classification <- classification.global
classification[classification.global == "Negative"] <- "Negative"
classification[classification.global == "Singlet"] <- hash.maxID[which(x = classification.global == "Singlet")]
classification[classification.global == "Doublet"] <- doublet_id[which(x = classification.global == "Doublet")]
classification.metadata <- data.frame(
hash.maxID,
hash.secondID,
hash.margin,
classification,
classification.global
)
colnames(x = classification.metadata) <- paste(
assay,
c('maxID', 'secondID', 'margin', 'classification', 'classification.global'),
sep = '_'
)
object <- AddMetaData(object = object, metadata = classification.metadata)
Idents(object) <- paste0(assay, '_classification')
# Idents(object, cells = rownames([email protected][[email protected]$classification.global == "Doublet", ])) <- "Doublet"
doublets <- rownames(x = object[[]])[which(object[[paste0(assay, "_classification.global")]] == "Doublet")]
Idents(object = object, cells = doublets) <- 'Doublet'
# [email protected]$hash.ID <- Idents(object)
object$hash.ID <- Idents(object = object)
return(object)
}
#' Calculate pearson residuals of features not in the scale.data
#'
#' This function calls sctransform::get_residuals.
#'
#' @param object A seurat object
#' @param features Name of features to add into the scale.data
#' @param assay Name of the assay of the seurat object generated by SCTransform
#' @param umi.assay Name of the assay of the seurat object containing UMI matrix and the default is
#' RNA
#' @param clip.range Numeric of length two specifying the min and max values the Pearson residual
#' will be clipped to
#' @param replace.value Recalculate residuals for all features, even if they are already present.
#' Useful if you want to change the clip.range.
#' @param verbose Whether to print messages and progress bars
#'
#' @return Returns a Seurat object containing pearson residuals of added features in its scale.data
#'
#' @importFrom sctransform get_residuals
#'
#' @export
#'
#' @seealso \code{\link[sctransform]{get_residuals}}
#'
#' @examples
#' pbmc_small <- SCTransform(object = pbmc_small, variable.features.n = 20)
#' pbmc_small <- GetResidual(object = pbmc_small, features = c('MS4A1', 'TCL1A'))
#'
GetResidual <- function(
object,
features,
assay = "SCT",
umi.assay = "RNA",
clip.range = NULL,
replace.value = FALSE,
verbose = TRUE
) {
if (!IsSCT(assay = object[[assay]])) {
stop(assay, " assay was not generated by SCTransform")
}
if (replace.value) {
new_features <- features
} else {
new_features <- setdiff(
x = features,
y = rownames(x = GetAssayData(object = object, assay = assay, slot = "scale.data"))
)
}
if (length(x = new_features) == 0) {
if (verbose) {
message("Pearson residuals of input features exist already")
}
} else {
if (is.null(x = Misc(object = object[[assay]], slot = 'vst.set'))) {
vst_out <- Misc(object = object[[assay]], slot = 'vst.out')
# filter cells not in the object but in the SCT model
vst_out$cell_attr <- vst_out$cell_attr[Cells(x = object), ]
vst_out$cells_step1 <- intersect(x = vst_out$cells_step1, y = Cells(x = object))
object <- GetResidualVstOut(
object = object,
assay = assay,
umi.assay = umi.assay,
new_features = new_features,
vst_out = vst_out,
clip.range = clip.range,
verbose = verbose
)
} else {
# Calculate Pearson Residual from integrated object SCT assay
vst.set <- Misc(object = object[[assay]], slot = 'vst.set')
scale.data <- GetAssayData(
object = object,
assay = assay,
slot = "scale.data"
)
vst_set_genes <- sapply(1:length(vst.set), function(x) rownames(vst.set[[x]]$model_pars_fit))
vst_set_genes <- Reduce(intersect, vst_set_genes)
diff_features <- setdiff(
x = new_features,
y = vst_set_genes
)
if (length(x = diff_features) !=0) {
warning(
"The following ", length(x = diff_features),
" features do not exist in all SCT models: ",
paste(diff_features, collapse = " ")
)
}
new_features <- intersect(
x = new_features,
y = vst_set_genes
)
if (length(new_features) != 0){
object <- SetAssayData(
object = object,
assay = assay,
slot = "scale.data",
new.data = scale.data[!rownames(x = scale.data) %in% new_features, , drop = FALSE]
)
new.scale.data <- matrix(nrow = length(new_features), ncol = 0)
rownames(x = new.scale.data) <- new_features
for (v in 1:length(x = vst.set)) {
vst_out <- vst.set[[v]]
# confirm that cells from SCT model also exist in the integrated object
cells.v <- intersect(x = rownames(x = vst_out$cell_attr), y = Cells(x = object))
vst_out$cell_attr <- vst_out$cell_attr[cells.v, ]
vst_out$cells_step1 <- intersect(x = vst_out$cells_step1, y = cells.v)
object.v <- subset(x = object, cells = cells.v)
object.v <- GetResidualVstOut(
object = object.v,
assay = assay,
umi.assay = umi.assay,
new_features = new_features,
vst_out = vst_out,
clip.range = clip.range,
verbose = verbose
)
new.scale.data <- cbind(
new.scale.data,
GetAssayData(object = object.v, assay = assay, slot ="scale.data" )[new_features, , drop = FALSE]
)
}
object <- SetAssayData(
object = object,
assay = assay,
slot = "scale.data",
new.data = rbind(
GetAssayData(object = object, slot = 'scale.data', assay = assay),
new.scale.data
)
)
}
}
}
return(object)
}
#' Normalize raw data
#'
#' Normalize count data per cell and transform to log scale
#'
#' @param data Matrix with the raw count data
#' @param scale.factor Scale the data. Default is 1e4
#' @param verbose Print progress
#'
#' @return Returns a matrix with the normalize and log transformed data
#'
#' @import Matrix
#' @importFrom methods as
#'
#' @export
#'
#' @examples
#' mat <- matrix(data = rbinom(n = 25, size = 5, prob = 0.2), nrow = 5)
#' mat
#' mat_norm <- LogNormalize(data = mat)
#' mat_norm
#'
LogNormalize <- function(data, scale.factor = 1e4, verbose = TRUE) {
if (is.data.frame(x = data)) {
data <- as.matrix(x = data)
}
if (!inherits(x = data, what = 'dgCMatrix')) {
data <- as(object = data, Class = "dgCMatrix")
}
# call Rcpp function to normalize
if (verbose) {
cat("Performing log-normalization\n", file = stderr())
}
norm.data <- LogNorm(data, scale_factor = scale.factor, display_progress = verbose)
colnames(x = norm.data) <- colnames(x = data)
rownames(x = norm.data) <- rownames(x = data)
return(norm.data)
}
#' Demultiplex samples based on classification method from MULTI-seq (McGinnis et al., bioRxiv 2018)
#'
#' Identify singlets, doublets and negative cells from multiplexing experiments. Annotate singlets by tags.
#'
#' @param object Seurat object. Assumes that the specified assay data has been added
#' @param assay Name of the multiplexing assay (HTO by default)
#' @param quantile The quantile to use for classification
#' @param autoThresh Whether to perform automated threshold finding to define the best quantile. Default is FALSE
#' @param maxiter Maximum number of iterations if autoThresh = TRUE. Default is 5
#' @param qrange A range of possible quantile values to try if autoThresh = TRUE
#' @param verbose Prints the output
#'
#' @return A Seurat object with demultiplexing results stored at \code{object$MULTI_ID}
#'
#' @import Matrix
#'
#' @export
#'
#' @references \url{https://www.biorxiv.org/content/early/2018/08/08/387241}
#'
#' @examples
#' \dontrun{
#' object <- MULTIseqDemux(object)
#' }
#'
MULTIseqDemux <- function(
object,
assay = "HTO",
quantile = 0.7,
autoThresh = FALSE,
maxiter = 5,
qrange = seq(from = 0.1, to = 0.9, by = 0.05),
verbose = TRUE
) {
assay <- assay %||% DefaultAssay(object = object)
multi_data_norm <- t(x = GetAssayData(
object = object,
slot = "data",
assay = assay
))
if (autoThresh) {
iter <- 1
negatives <- c()
neg.vector <- c()
while (iter <= maxiter) {
# Iterate over q values to find ideal barcode thresholding results by maximizing singlet classifications
bar.table_sweep.list <- list()
n <- 0
for (q in qrange) {
n <- n + 1
# Generate list of singlet/doublet/negative classifications across q sweep
bar.table_sweep.list[[n]] <- ClassifyCells(data = multi_data_norm, q = q)
names(x = bar.table_sweep.list)[n] <- paste0("q=" , q)
}
# Determine which q values results in the highest pSinglet
res_round <- FindThresh(call.list = bar.table_sweep.list)$res
res.use <- res_round[res_round$Subset == "pSinglet", ]
q.use <- res.use[which.max(res.use$Proportion),"q"]
if (verbose) {
message("Iteration ", iter)
message("Using quantile ", q.use)
}
round.calls <- ClassifyCells(data = multi_data_norm, q = q.use)
#remove negative cells
neg.cells <- names(x = round.calls)[which(x = round.calls == "Negative")]
neg.vector <- c(neg.vector, rep(x = "Negative", length(x = neg.cells)))
negatives <- c(negatives, neg.cells)
if (length(x = neg.cells) == 0) {
break
}
multi_data_norm <- multi_data_norm[-which(x = rownames(x = multi_data_norm) %in% neg.cells), ]
iter <- iter + 1
}
names(x = neg.vector) <- negatives
demux_result <- c(round.calls,neg.vector)
demux_result <- demux_result[rownames(x = object[[]])]
} else{
demux_result <- ClassifyCells(data = multi_data_norm, q = quantile)
}
demux_result <- demux_result[rownames(x = object[[]])]
object[['MULTI_ID']] <- factor(x = demux_result)
Idents(object = object) <- "MULTI_ID"
bcs <- colnames(x = multi_data_norm)
bc.max <- bcs[apply(X = multi_data_norm, MARGIN = 1, FUN = which.max)]
bc.second <- bcs[unlist(x = apply(
X = multi_data_norm,
MARGIN = 1,
FUN = function(x) {
return(which(x == MaxN(x)))
}
))]
doublet.names <- unlist(x = lapply(
X = 1:length(x = bc.max),
FUN = function(x) {
return(paste(sort(x = c(bc.max[x], bc.second[x])), collapse = "_"))
}
))
doublet.id <- which(x = demux_result == "Doublet")
MULTI_classification <- as.character(object$MULTI_ID)
MULTI_classification[doublet.id] <- doublet.names[doublet.id]
object$MULTI_classification <- factor(x = MULTI_classification)
return(object)
}
#' Load in data from Alevin pipeline
#'
#' Enables easy loading of csv format matrix provided by Alevin
#' ran with `--dumpCsvCounts` flags.
#'
#' @param base.path Directory containing the alevin/quant_mat*
#' files provided by Alevin.
#'
#' @return Returns a matrix with rows and columns labeled
#'
#' @importFrom utils read.csv read.delim
#' @export
#'
#' @author Avi Srivastava
#'
#' @examples
#' \dontrun{
#' data_dir <- 'path/to/output/directory'
#' list.files(data_dir) # Should show alevin/quants_mat* files
#' expression_matrix <- ReadAlevinCsv(base.path = data_dir)
#' seurat_object = CreateSeuratObject(counts = expression_matrix)
#' }
#'
ReadAlevinCsv <- function(base.path) {
if (!dir.exists(base.path)) {
stop("Directory provided does not exist")
}
barcode.loc <- file.path(base.path, "alevin", "quants_mat_rows.txt")
gene.loc <- file.path(base.path, "alevin", "quants_mat_cols.txt")
matrix.loc <- file.path( base.path, "alevin", "quants_mat.csv" )
if (!file.exists(barcode.loc)) {
stop("Barcode file missing")
}
if (!file.exists(gene.loc)) {
stop("Gene name file missing")
}
if (!file.exists(matrix.loc)) {
stop("Expression matrix file missing")
}
matrix <- as.matrix(x = read.csv(file = matrix.loc, header = FALSE))
matrix <- t(x = matrix[, 1:ncol(x = matrix) - 1])
cell.names <- readLines(con = barcode.loc)
gene.names <- readLines(con = gene.loc)
colnames(x = matrix) <- cell.names
rownames(x = matrix) <- gene.names
matrix[is.na(x = matrix)] <- 0
return(matrix)
}
#' Load in data from Alevin pipeline
#'
#' Enables easy loading of binary format matrix provided by Alevin
#'
#' @param base.path Directory containing the alevin/quant_mat*
#' files provided by Alevin.
#'
#' @return Returns a matrix with rows and columns labeled
#'
#' @export
#'
#' @author Avi Srivastava
#'
#' @examples
#' \dontrun{
#' data_dir <- 'path/to/output/directory'
#' list.files(data_dir) # Should show alevin/quants_mat* files
#' expression_matrix <- ReadAlevin(base.path = data_dir)
#' seurat_object = CreateSeuratObject(counts = expression_matrix)
#' }
#'
ReadAlevin <- function(base.path) {
if (!dir.exists(base.path)) {
stop("Directory provided does not exist")
}
barcode.loc <- file.path(base.path, "alevin", "quants_mat_rows.txt")
gene.loc <- file.path(base.path, "alevin", "quants_mat_cols.txt")
matrix.loc <- file.path(base.path, "alevin", "quants_mat.gz")
if (!file.exists(barcode.loc)) {
stop("Barcode file missing")
}
if (!file.exists(gene.loc)) {
stop("Gene name file missing")
}
if (!file.exists(matrix.loc)) {
stop("Expression matrix file missing")
}
cell.names <- readLines(con = barcode.loc)
gene.names <- readLines(con = gene.loc)
num.cells <- length(x = cell.names)
num.genes <- length(x = gene.names)
out.matrix <- matrix(data = NA, nrow = num.genes, ncol = num.cells)
con <- gzcon(con = file(description = matrix.loc, open = "rb"))
total.molecules <- 0.0
for (n in seq_len(length.out = num.cells)) {
out.matrix[, n] <- readBin(
con = con,
what = double(),
endian = "little",
n = num.genes
)
total.molecules <- total.molecules + sum(out.matrix[, n])
}
colnames(x = out.matrix) <- cell.names
rownames(x = out.matrix) <- gene.names
message("Found total ", total.molecules, " molecules")
return(out.matrix)
}
#' Load in data from 10X
#'
#' Enables easy loading of sparse data matrices provided by 10X genomics.
#'
#' @param data.dir Directory containing the matrix.mtx, genes.tsv (or features.tsv), and barcodes.tsv
#' files provided by 10X. A vector or named vector can be given in order to load
#' several data directories. If a named vector is given, the cell barcode names
#' will be prefixed with the name.
#' @param gene.column Specify which column of genes.tsv or features.tsv to use for gene names; default is 2
#' @param unique.features Make feature names unique (default TRUE)
#'
#' @return If features.csv indicates the data has multiple data types, a list
#' containing a sparse matrix of the data from each type will be returned.
#' Otherwise a sparse matrix containing the expression data will be returned.
#'
#' @importFrom Matrix readMM
#'
#' @export
#'
#' @examples
#' \dontrun{
#' # For output from CellRanger < 3.0
#' data_dir <- 'path/to/data/directory'
#' list.files(data_dir) # Should show barcodes.tsv, genes.tsv, and matrix.mtx
#' expression_matrix <- Read10X(data.dir = data_dir)
#' seurat_object = CreateSeuratObject(counts = expression_matrix)
#'
#' # For output from CellRanger >= 3.0 with multiple data types
#' data_dir <- 'path/to/data/directory'
#' list.files(data_dir) # Should show barcodes.tsv.gz, features.tsv.gz, and matrix.mtx.gz
#' data <- Read10X(data.dir = data_dir)
#' seurat_object = CreateSeuratObject(counts = data$`Gene Expression`)
#' seurat_object[['Protein']] = CreateAssayObject(counts = data$`Antibody Capture`)
#' }
#'
Read10X <- function(data.dir = NULL, gene.column = 2, unique.features = TRUE) {
full.data <- list()
for (i in seq_along(along.with = data.dir)) {
run <- data.dir[i]
if (!dir.exists(paths = run)) {
stop("Directory provided does not exist")
}
barcode.loc <- file.path(run, 'barcodes.tsv')
gene.loc <- file.path(run, 'genes.tsv')
features.loc <- file.path(run, 'features.tsv.gz')
matrix.loc <- file.path(run, 'matrix.mtx')
# Flag to indicate if this data is from CellRanger >= 3.0
pre_ver_3 <- file.exists(gene.loc)
if (!pre_ver_3) {
addgz <- function(s) {
return(paste0(s, ".gz"))
}
barcode.loc <- addgz(s = barcode.loc)
matrix.loc <- addgz(s = matrix.loc)
}
if (!file.exists(barcode.loc)) {
stop("Barcode file missing. Expecting ", basename(path = barcode.loc))
}
if (!pre_ver_3 && !file.exists(features.loc) ) {
stop("Gene name or features file missing. Expecting ", basename(path = features.loc))
}
if (!file.exists(matrix.loc)) {
stop("Expression matrix file missing. Expecting ", basename(path = matrix.loc))
}
data <- readMM(file = matrix.loc)
cell.names <- readLines(barcode.loc)
if (all(grepl(pattern = "\\-1$", x = cell.names))) {
cell.names <- as.vector(x = as.character(x = sapply(
X = cell.names,
FUN = ExtractField,
field = 1,
delim = "-"
)))
}
if (is.null(x = names(x = data.dir))) {
if (i < 2) {
colnames(x = data) <- cell.names
} else {
colnames(x = data) <- paste0(i, "_", cell.names)
}
} else {
colnames(x = data) <- paste0(names(x = data.dir)[i], "_", cell.names)
}
feature.names <- read.delim(
file = ifelse(test = pre_ver_3, yes = gene.loc, no = features.loc),
header = FALSE,
stringsAsFactors = FALSE
)
if (any(is.na(x = feature.names[, gene.column]))) {
warning(
'Some features names are NA. Replacing NA names with ID from the opposite column requested',
call. = FALSE,
immediate. = TRUE
)
na.features <- which(x = is.na(x = feature.names[, gene.column]))
replacement.column <- ifelse(test = gene.column == 2, yes = 1, no = 2)
feature.names[na.features, gene.column] <- feature.names[na.features, replacement.column]
}
if (unique.features) {
fcols = ncol(x = feature.names)
if (fcols < gene.column) {
stop(paste0("gene.column was set to ", gene.column,
" but feature.tsv.gz (or genes.tsv) only has ", fcols, " columns.",
" Try setting the gene.column argument to a value <= to ", fcols, "."))
}
rownames(x = data) <- make.unique(names = feature.names[, gene.column])
}
# In cell ranger 3.0, a third column specifying the type of data was added
# and we will return each type of data as a separate matrix
if (ncol(x = feature.names) > 2) {
data_types <- factor(x = feature.names$V3)
lvls <- levels(x = data_types)
if (length(x = lvls) > 1 && length(x = full.data) == 0) {
message("10X data contains more than one type and is being returned as a list containing matrices of each type.")
}
expr_name <- "Gene Expression"
if (expr_name %in% lvls) { # Return Gene Expression first
lvls <- c(expr_name, lvls[-which(x = lvls == expr_name)])
}
data <- lapply(
X = lvls,
FUN = function(l) {
return(data[data_types == l, ])
}
)
names(x = data) <- lvls
} else{
data <- list(data)
}
full.data[[length(x = full.data) + 1]] <- data
}
# Combine all the data from different directories into one big matrix, note this
# assumes that all data directories essentially have the same features files
list_of_data <- list()
for (j in 1:length(x = full.data[[1]])) {
list_of_data[[j]] <- do.call(cbind, lapply(X = full.data, FUN = `[[`, j))
# Fix for Issue #913
list_of_data[[j]] <- as(object = list_of_data[[j]], Class = "dgCMatrix")
}
names(x = list_of_data) <- names(x = full.data[[1]])
# If multiple features, will return a list, otherwise
# a matrix.
if (length(x = list_of_data) == 1) {
return(list_of_data[[1]])
} else {
return(list_of_data)
}
}
#' Read 10X hdf5 file
#'
#' Read count matrix from 10X CellRanger hdf5 file.
#' This can be used to read both scATAC-seq and scRNA-seq matrices.
#'
#' @param filename Path to h5 file
#' @param use.names Label row names with feature names rather than ID numbers.
#' @param unique.features Make feature names unique (default TRUE)
#'
#' @return Returns a sparse matrix with rows and columns labeled. If multiple
#' genomes are present, returns a list of sparse matrices (one per genome).
#'
#' @export
#'
Read10X_h5 <- function(filename, use.names = TRUE, unique.features = TRUE) {
if (!requireNamespace('hdf5r', quietly = TRUE)) {
stop("Please install hdf5r to read HDF5 files")
}
if (!file.exists(filename)) {
stop("File not found")
}
infile <- hdf5r::H5File$new(filename = filename, mode = 'r')
genomes <- names(x = infile)
output <- list()
if (!infile$attr_exists("PYTABLES_FORMAT_VERSION")) {
# cellranger version 3
if (use.names) {
feature_slot <- 'features/name'
} else {
feature_slot <- 'features/id'
}
} else {
if (use.names) {
feature_slot <- 'gene_names'
} else {
feature_slot <- 'genes'
}
}
for (genome in genomes) {