-
Notifications
You must be signed in to change notification settings - Fork 0
/
snATAC_processing.R
201 lines (180 loc) · 9.08 KB
/
snATAC_processing.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
library(Seurat)
library(Signac)
library(GenomeInfoDb)
library(tidyverse)
################################################################################
# Step 01: Load 10X aggregated peaks by cells matrix into seurat
################################################################################
peaks <- Read10X_h5("cellranger_aggr_dir/outs/filtered_peak_bc_matrix.h5")
cellranger_meta <- read.csv("cellranger_aggr/outs/singlecell.csv", header=T, row.names=1)
NucSeq.atac <- CreateSeuratObject(
counts=peaks,
assay="peaks",
meta.data=cellranger_meta,
min.cells=1
)
fragment.path <- "cellranger_aggr/outs/fragments.tsv.gz"
NucSeq.atac <- SetFragments(NucSeq.atac, file = fragment.path)
# binarize peaks
NucSeq.atac@assays$peaks@counts@x[NucSeq.atac@assays$peaks@counts@x > 0] <- 1
# add sample level metadata (Diagnosis, Age, Sex, etc)
my.data <- read.csv(file = "data/CellSampleID.csv", header=TRUE, row.names=1)
all.equal(rownames(my.data), colnames(NucSeq.atac))
NucSeq.atac <- AddMetaData(NucSeq.atac, metadata=my.data)
################################################################################
# Step 02: Quality Control
################################################################################
NucSeq.atac$pct_reads_in_peaks <- NucSeq.atac$peak_region_fragments / NucSeq.atac$nCount_peaks * 100
NucSeq.atac$blacklist_ratio <- NucSeq.atac$blacklist_region_fragments / NucSeq.atac$peak_region_fragments
# for each sample, compute nucleosome banding
# note: download chromosome sizes file from UCSC https://hgdownload-test.gi.ucsc.edu/goldenPath/hg38/bigZips/
chrom.sizes = read.csv('hg38.chrom.sizes', sep='\t', header=F)
colnames(chrom.sizes) <- c('chr', 'size')
NucSeq.atac$nucleosome_signal <- NA
NucSeq.atac$nucleosome_group <- NA
samples <- unique(NucSeq.atac$Sample.ID)
for(i in 1:length(samples)){
print(samples[i])
temp <- NucleosomeSignal(
subset(NucSeq.atac, Sample.ID == samples[i]),
region=paste0(chrom.sizes[1,][[1]],"-1-",chrom.sizes[1,][[2]])
)
temp$nucleosome_group <- ifelse(temp$nucleosome_signal > 10, 'NS > 10', 'NS < 10')
NucSeq.atac$nucleosome_signal <- ifelse(NucSeq.atac$Sample.ID == samples[i], temp$nucleosome_signal, NucSeq.atac$nucleosome_signal)
NucSeq.atac$nucleosome_group <- ifelse(NucSeq.atac$Sample.ID == samples[i], temp$nucleosome_group, NucSeq.atac$nucleosome_group)
}
NucSeq.atac$pass_qc <- ifelse(NucSeq.atac$peak_region_fragments > 300 & NucSeq.atac$peak_region_fragments < 10000 & NucSeq.atac$pct_reads_in_peaks > 15 & NucSeq.atac$blacklist_ratio < 0.01 & NucSeq.atac$nucleosome_signal < 10, TRUE, FALSE)
NucSeq.atac <- NucSeq.atac[,NucSeq.atac$pass_qc]
################################################################################
# Step 03: Primary Processing
################################################################################
# option 1: Process with Seurat / Signac (not used)
NucSeq.atac <- RunTFIDF(NucSeq.atac)
NucSeq.atac <- RunSVD(
object = NucSeq.atac,
assay = 'peaks',
reduction.key = 'LSI_',
reduction.name = 'lsi'
)
NucSeq.atac <- RunUMAP(NucSeq.atac, reduction='lsi', dims=1:30)
# option 2: Process with monocle3
peak_matrix <- GetAssayData(NucSeq.atac, slot='counts')
peaks <- data.frame(as.character(rownames(peak_matrix)))
rownames(peaks) <- rownames(peak_matrix)
peaks <- as.data.frame(cbind(peaks,peaks))
colnames(peaks) <- c("GeneSymbol", "gene_short_name")
NucSeq.atac_cds <- new_cell_data_set(
peak_matrix,
cell_metadata=NucSeq.atac@meta.data,
gene_metadata=genes
)
NucSeq.atac_cds <- detect_genes(NucSeq.atac_cds)
NucSeq.atac_cds <- estimate_size_factors(NucSeq.atac_cds)
NucSeq.atac_cds <- preprocess_cds(NucSeq.atac_cds, method = "LSI")
NucSeq.atac_cds <- align_cds(NucSeq.atac_cds, preprocess_method='LSI', alignment_group = "Batch")
NucSeq.atac_cds <- reduce_dimension(NucSeq.atac_cds, reduction_method = 'UMAP', preprocess_method = "LSI")
NucSeq.atac_cds <- cluster_cells(NucSeq.atac_cds)
monocle_umap <- NucSeq.atac_cds@reducedDims[["UMAP"]][rownames(NucSeq.atac_cds@reducedDims[["UMAP"]]) %in% colnames(NucSeq.atac),]
colnames(monocle_umap) <- c('UMAP_1', 'UMAP_2')
all.equal(rownames(monocle_umap), colnames(NucSeq.atac))
NucSeq.atac@reductions$umap@cell.embeddings <- monocle_umap
NucSeq.atac$monocle_clusters_umap <- clusters(NucSeq.atac_cds, reduction_method='UMAP')
monocle_aligned <- NucSeq.atac_cds@reducedDims[["Aligned"]]
rownames(monocle_aligned) <- colnames(NucSeq.atac_cds)
colnames(monocle_aligned) <- paste0('LSI_', seq(1:ncol(monocle_aligned)))
all.equal(rownames(monocle_aligned), colnames(NucSeq.atac))
NucSeq.atac@reductions$lsi <- CreateDimReducObject(
embeddings=monocle_aligned,
key="LSI_",
assay="peaks"
)
################################################################################
# Step 04: Construct gene activity matrix
################################################################################
library(EnsDb.Hsapiens.v86)
gene.coords <- genes(EnsDb.Hsapiens.v86, filter = ~ gene_biotype == "protein_coding")
seqlevelsStyle(gene.coords) <- 'UCSC'
genebody.coords <- keepStandardChromosomes(gene.coords, pruning.mode = 'coarse')
genebodyandpromoter.coords <- Extend(x = gene.coords, upstream = 2000, downstream = 0)
gene.activities <- FeatureMatrix(
fragments = fragment.path,
features = genebodyandpromoter.coords,
cells = colnames(NucSeq.atac),
chunk = 10
)
gene.key <- genebodyandpromoter.coords$gene_name
names(gene.key) <- GRangesToString(grange = genebodyandpromoter.coords)
rownames(gene.activities) <- gene.key[rownames(gene.activities)]
NucSeq.atac[['RNA']] <- CreateAssayObject(counts = gene.activities)
NucSeq.atac <- NormalizeData(
object = NucSeq.atac,
assay = 'RNA',
normalization.method = 'LogNormalize',
scale.factor = median(NucSeq.atac$nCount_RNA)
)
DefaultAssay(NucSeq.atac) <- 'RNA'
################################################################################
# Step 04: Construct TF activity matrix (warning: takes a lot of time/memory!!!)
################################################################################
library(JASPAR2018)
library(TFBSTools)
library(BSgenome.Hsapiens.UCSC.hg38)
pfm <- getMatrixSet(x = JASPAR2018,opts = list(species = 9606, all_versions = FALSE))
motif.matrix <- CreateMotifMatrix(features = StringToGRanges(rownames(NucSeq.atac), sep = c(":", "-")),pwm = pfm,genome = 'hg38',sep = c(":", "-"))
motif <- CreateMotifObject(data = motif.matrix,pwm = pfm)
NucSeq.atac[['peaks']] <- AddMotifObject(object = NucSeq.atac[['peaks']],motif.object = motif)
NucSeq.atac <- RunChromVAR(object = NucSeq.atac,genome = BSgenome.Hsapiens.UCSC.hg38)
################################################################################
# Step 05: integrate with snRNA-seq data
################################################################################
NucSeq.rna <- readRDS('data/snRNA.rds')
NucSeq.rna$tech <- 'rna'; NucSeq.atac$tech <- 'atac';
DefaultAssay(NucSeq.atac) <- 'RNA'
# compute anchors between RNA and ATAC
transfer.anchors <- FindTransferAnchors(
reference=NucSeq.rna,
query=NucSeq.atac,
features=VariableFeatures(NucSeq.rna),
reference.assay="RNA",
query.assay="RNA",
reduction="cca",
verbose=T,
dims=1:40
)
celltype.predictions <- TransferData(
anchorset=transfer.anchors,
refdata=NucSeq.rna$Cell.Type,
weight.reduction=NucSeq.atac[["lsi"]]
)
NucSeq.atac <- AddMetaData(NucSeq.atac, celltype.predictions)
NucSeq.atac$predicted.id <- factor(NucSeq.atac$predicted.id, levels = levels(as.factor(NucSeq.rna$Cell.Type)))
Idents(NucSeq.atac) <- NucSeq.atac$monocle_clusters_umap_Cell.Type
Idents(NucSeq.rna) <- NucSeq.rna$Cell.Type
Idents(NucSeq.atac) <- paste0(Idents(NucSeq.atac), 'atac')
Idents(NucSeq.rna) <- paste0(Idents(NucSeq.rna), 'rna')
genes.use <- VariableFeatures(NucSeq.rna)
refdata <- GetAssayData(NucSeq.rna, assay = "RNA", slot = "data")[genes.use, ]
imputation <- TransferData(anchorset = transfer.anchors, refdata = refdata, weight.reduction = NucSeq.atac[["lsi"]])
NucSeq.atac[["RNA"]] <- imputation
NucSeq.atac <- RenameCells(NucSeq.atac, add.cell.id='atac')
NucSeq.rna <- RenameCells(NucSeq.rna, add.cell.id='rna')
NucSeq.coembed <- merge(x = NucSeq.rna, y = NucSeq.atac)
cells_to_keep <- c(colnames(NucSeq.rna), colnames(NucSeq.atac)[NucSeq.atac$prediction.score.max >= 0.5])
NucSeq.coembed <- NucSeq.coembed[,cells_to_keep]
NucSeq.coembed <- ScaleData(NucSeq.coembed, features = genes.use, do.scale = FALSE)
NucSeq.coembed <- RunPCA(NucSeq.coembed, features = rownames(NucSeq.coembed), verbose = FALSE)
NucSeq.coembed <- RunUMAP(NucSeq.coembed, dims = 1:30)
expr_matrix <- GetAssayData(NucSeq.coembed, slot='data', assay='RNA')
genes <- data.frame(as.character(rownames(expr_matrix)))
rownames(genes) <- rownames(expr_matrix)
genes <- as.data.frame(cbind(genes,genes))
colnames(genes) <- c("GeneSymbol", "gene_short_name")
NucSeq_cds <- new_cell_data_set(
expr_matrix,
cell_metadata=NucSeq.coembed@meta.data,
gene_metadata=genes
)
NucSeq_cds@reducedDims[['PCA']] <- NucSeq.coembed@reductions$pca@cell.embeddings
NucSeq_cds <- align_cds(NucSeq_cds, preprocess_method='PCA', alignment_group = "Batch")
NucSeq_cds <- reduce_dimension(NucSeq_cds, reduction_method = 'UMAP', preprocess_method = "Aligned")
NucSeq_cds <- cluster_cells(NucSeq_cds, reduction_method='UMAP')