sc-TissueMapper_functions.R 43.6 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
#sc-TissueMapper
#Author: Gervaise H. Henry
#Email: gervaise.henry@utsouthwestern.edu
#Lab: Strand Lab, Deparment of Urology, University of Texas Southwestern Medical Center


scFolders <- function(){
  if (!dir.exists("./analysis/qc/")){
    dir.create("./analysis/qc/")
  }
11
12
13
14
15
16
  if (!dir.exists("./analysis/qc/")){
    dir.create("./analysis/qc/")
  }
  if (!dir.exists("./analysis/qc/cutoffs/")){
    dir.create("./analysis/qc/cutoffs/")
  }
17
18
19
20
21
22
23
24
25
26
27
28
  if (!dir.exists("./analysis/qc/cellcycle")){
    dir.create("./analysis/qc/cellcycle")
  }
  if (!dir.exists("./analysis/vis")){
    dir.create("./analysis/vis")
  }
  if (!dir.exists("./analysis/score_id")){
    dir.create("./analysis/score_id")
  }
  if (!dir.exists("./analysis/cor")){
    dir.create("./analysis/cor")
  }
29
30
31
32
33
34
  if (!dir.exists("./analysis/shiny")){
    dir.create("./analysis/shiny")
  }
  if (!dir.exists("./analysis/shiny")){
    dir.create("./analysis/shiny")
  }
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
  for (i in c("raw","id","id.epi","id.fmst","id.st","id.leu")){
    if (!dir.exists(paste0("./analysis/shiny/",i))){
      dir.create(paste0("./analysis/shiny/",i))
    }
    if (!dir.exists(paste0("./analysis/shiny/",i,"/outs"))){
      dir.create(paste0("./analysis/shiny/",i,"/outs"))
    }
    if (!dir.exists(paste0("./analysis/shiny/",i,"/outs/filtered_feature_bc_matrix"))){
      dir.create(paste0("./analysis/shiny/",i,"/outs/filtered_feature_bc_matrix"))
    }
    if (!dir.exists(paste0("./analysis/shiny/",i,"/outs/analysis"))){
      dir.create(paste0("./analysis/shiny/",i,"/outs/analysis"))
    }
    if (!dir.exists(paste0("./analysis/shiny/",i,"/outs/analysis/clustering"))){
      dir.create(paste0("./analysis/shiny/",i,"/outs/analysis/clustering"))
    }
    if (!dir.exists(paste0("./analysis/shiny/",i,"/outs/analysis/diffexp"))){
      dir.create(paste0("./analysis/shiny/",i,"/outs/analysis/diffexp"))
    }
    if (!dir.exists(paste0("./analysis/shiny/",i,"/outs/analysis/pca"))){
      dir.create(paste0("./analysis/shiny/",i,"/outs/analysis/pca"))
    }
    if (!dir.exists(paste0("./analysis/shiny/",i,"/outs/analysis/pca/10_components"))){
      dir.create(paste0("./analysis/shiny/",i,"/outs/analysis/pca/10_components"))
    }
    if (!dir.exists(paste0("./analysis/shiny/",i,"/outs/analysis/tsne"))){
      dir.create(paste0("./analysis/shiny/",i,"/outs/analysis/tsne"))
    }
    if (!dir.exists(paste0("./analysis/shiny/",i,"/outs/analysis/tsne/2_components"))){
      dir.create(paste0("./analysis/shiny/",i,"/outs/analysis/tsne/2_components"))
    }
    if (!dir.exists(paste0("./analysis/shiny/",i,"/outs/analysis/umap"))){
      dir.create(paste0("./analysis/shiny/",i,"/outs/analysis/umap"))
    }
    if (!dir.exists(paste0("./analysis/shiny/",i,"/outs/analysis/umap/2_components"))){
      dir.create(paste0("./analysis/shiny/",i,"/outs/analysis/umap/2_components"))
    }
72
  }
73
74
75
}


Gervaise Henry's avatar
Gervaise Henry committed
76
scLoad <- function(p,cellranger=3,aggr=TRUE,ncell=0,nfeat=0){
77
78
79
80
  #Load and prefilter filtered_gene_bc_matrices_mex output from cellranger
  
  #Inputs:
  #p = project name
81
82
  #cellranger cellranger version number used for count/aggr, 2 or 3
  #aggr = if the samples are already aggregated, TRUE if useing the output of aggr, FALSE if using outputs of each count
83
84
  
  #Outputs:
85
86
  #sc10x = Seurat object list
  #sc10x.groups = group labels for each sample
87
88
  
  
89
90
91
92
93
94
95
  sc10x.groups <- read_csv(paste0("./analysis/DATA/",p,"-demultiplex.csv"))
  
  
  #Load filtered_gene_bc_matrices output from cellranger
  sc10x.data <- list()
  sc10x <- list()
  if (aggr==TRUE){
96
    if (cellranger==2){
97
      sc10x.data[aggr] <- Read10X(data.dir=paste0("./analysis/DATA/10x/filtered_gene_bc_matrices_mex/"))
98
    } else {
99
      sc10x.data[aggr] <- Read10X(data.dir=paste0("./analysis/DATA/10x/filtered_feature_bc_matrix/"))
100
    }
101
    sc10x[aggr] <- new("seurat",raw.data=sc10x.data[aggr])
102
  } else {
103
104
105
106
107
108
    for (i in sc10x.groups$Samples){
      if (cellranger==2){
        sc10x.data[i] <- Read10X(data.dir=paste0("./analysis/DATA/10x/",i,"/filtered_gene_bc_matrices/"))
      } else {
        sc10x.data[i] <- Read10X(data.dir=paste0("./analysis/DATA/10x/",i,"/filtered_feature_bc_matrix/"))
      }
Gervaise Henry's avatar
Gervaise Henry committed
109
      sc10x[i] <- CreateSeuratObject(counts=sc10x.data[[i]],project=p,min.cells=ncell,min.features=nfeat)
110
      sc10x[[i]]$samples <- i
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
  # #Label sample names from aggregation_csv.csv
  # if (sub==FALSE){
  #   if (cellranger==2){
  #     sc10x.aggr <- read_csv("./analysis/DATA/10x/aggregation_csv.csv")
  #   } else {
  #     sc10x.aggr <- read_csv("./analysis/DATA/10x/aggregation.csv")
  #   }
  # } else {
  #   if (cellranger==2){
  #     sc10x.aggr <- read_csv(paste0("./analysis/DATA/",p,"/10x/aggregation_csv.csv"))
  #   } else {
  #     sc10x.aggr <- read_csv(paste0("./analysis/DATA/",p,"/10x/aggregation.csv"))
  #   }
  # }
  # cell.codes <- as.data.frame(sc10x@raw.data@Dimnames[[2]])
  # colnames(cell.codes) <- "barcodes"
  # rownames(cell.codes) <- cell.codes$barcodes
  # cell.codes$lib.codes <- as.factor(gsub(pattern=".+-",replacement="",cell.codes$barcodes))
  # cell.codes$samples <- sc10x.aggr$library_id[match(cell.codes$lib.codes,as.numeric(rownames(sc10x.aggr)))]
  # sc10x <- CreateSeuratObject(counts=sc10x.data,project=p,assay="RNA",min.cells=mc,min.features=mg,meta.data=cell.codes["samples"])
  # 
  # #Create groups found in demultiplex.csv
  # for (i in 2:ncol(sc10x.demultiplex)){
  #   Idents(sc10x) <- "samples"
  #   merge.cluster <- apply(sc10x.demultiplex[,i],1,as.character)
  #   merge.cluster[merge.cluster==1] <- colnames(sc10x.demultiplex[,i])
  #   
  #   Idents(sc10x) <- plyr::mapvalues(x=Idents(sc10x),from=sc10x.demultiplex$Samples,to=merge.cluster)
  #   sc10x@meta.data[,colnames(sc10x.demultiplex[,i])] <- Idents(sc10x)
  # }
  
  
  results <- list(
    sc10x=sc10x,
    sc10x.groups=sc10x.groups
  )
  return(results)
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
}


scSubset <- function(sc10x,i="ALL",g="ALL"){
  #Subset cells based on an identity
  
  #Inputs:
  #sc10x = seruat object
  #i = identity to use
  #g = group to subset by
  
  #Outputs:
  #Seurat object
  
  
166
  Idents(sc10x) <- i
167
168
  sc10x.sub <- subset(x=sc10x,idents=g)
  
169
  
170
171
172
173
  return(sc10x.sub)
}


174
scQC <- function(sc10x,sp="hu",feature="nFeature_RNA"){
175
176
177
178
179
180
181
  #QC and filter Seurat object
  
  #Inputs:
  #sc10x = Seruat object
  #sub = Subfolder to save output files
  
  #Outputs:
182
  #result[1] = filtered Seurat object
183
184
185
186
187
188
  #result[2] = raw cell count
  #result[3] = raw gene count
  #result[4] = filtered cell count
  #result[5] = filtered gene count
  
  
189
  #Calculate percent mitochondrea
190
  if (sp=="hu"){
191
    mito.pattern <- "^MT-"
192
    ribo.pattern <- "^(RPL|RPS)"
193
194
  } else if (sp=="mu"){
    mito.pattern <- "^mt-"
195
    ribo.pattern <- "^(Rpl|Rps)"
196
  }
197
198
199
  for (i in names(sc10x)){
    sc10x.temp <- sc10x[[i]]
    sc10x.temp[["percent.mito"]] <- PercentageFeatureSet(object=sc10x.temp,pattern=mito.pattern)
200
    sc10x.temp[["percent.ribo"]] <- PercentageFeatureSet(object=sc10x.temp,pattern=ribo.pattern)
201
    #sc10x.temp <- subset(sc10x.temp,cell=names(which(is.na(sc10x.temp$percent.mito))),invert=TRUE)
202
    sc10x[i] <- sc10x.temp
203
  }
204
  
205
206
  #Calculate cutoffs
  thresh <- list()
Gervaise Henry's avatar
Gervaise Henry committed
207
208
209
  #h <- list()
  #cells.remove <- list()
  #sc10x.temp <- list()
210
  for (i in feature){
Gervaise Henry's avatar
Gervaise Henry committed
211
    if (i == "nFeature_RNA"){
212
213
214
215
216
217
218
219
220
      h <- list()
      cells.remove <- list()
      sc10x.temp <- list()
      for (j in names(sc10x)){
        h[[i]] <- hist(data.frame(sc10x[[j]][[i]])$nFeature_RNA,breaks=10,plot=FALSE)
        cutoff.temp <- mean(c(h[[i]]$mids[which.max(h[[i]]$counts)],h[[i]]$mids[-which.max(h[[i]]$counts)][which.max(h[[i]]$counts[-which.max(h[[i]]$counts)])]))
        cells.remove[[j]] <- c(cells.remove[[j]],rownames(sc10x[[j]][["nFeature_RNA"]])[sc10x[[j]][[i]][,1] < cutoff.temp])
        sc10x.temp[[j]] <- subset(sc10x[[j]],cells=setdiff(colnames(sc10x[[j]]),cells.remove[[j]]))
      }
Gervaise Henry's avatar
Gervaise Henry committed
221
222
    thresh[[i]] <- scThresh(sc10x.temp,feature=i)
    }
223
    if (i == "percent.mito"){
224
      thresh[[i]] <- scThresh(sc10x,feature=i)
Gervaise Henry's avatar
Gervaise Henry committed
225
226
227
228
229
      h <- list()
      cells.remove <- list()
      sc10x.temp <- list()
      thresh.l <- list()
      cutoff.l.mito <- list()
230
231
      for (j in names(sc10x)){
        cutoff.l.mito[[j]] <- NULL
232
233
        temp <- data.frame(sc10x[[j]][[i]])$percent.mito
        h[[i]] <- hist(temp,breaks=100,plot=FALSE)
234
235
        cutoff.temp <- mean(c(h[[i]]$mids[which.max(h[[i]]$counts)],h[[i]]$mids[-which.max(h[[i]]$counts)][which.max(h[[i]]$counts[-which.max(h[[i]]$counts)])]))
        cells.remove[[j]] <- c(cells.remove[[j]],rownames(sc10x[[j]][["percent.mito"]])[sc10x[[j]][[i]][,1] > cutoff.temp])
236
        sc10x.temp[[j]] <- subset(sc10x[[j]],cells=setdiff(colnames(sc10x[[j]]),cells.remove[[j]]))
237
        thresh.l[[i]] <- scThresh(sc10x.temp,feature=i,sub="lower")
Gervaise Henry's avatar
Gervaise Henry committed
238
239
        #cutoff.l.mito[[j]] <- thresh.l[[i]][[j]]$threshold[thresh.l[[i]][[j]]$method=="Triangle"]
        cutoff.l.mito[[j]] <- thresh.l[[i]][[j]]$threshold[thresh.l[[i]][[j]]$method=="RenyiEntropy"]
240
      }
241
    }
242
    if (i == "percent.ribo"){
243
244
245
246
      thresh[[i]] <- scThresh(sc10x,feature=i)
    }
    if (i == "nCount_RNA"){
      thresh[[i]] <- scThresh(sc10x,feature=i)
247
248
249
250
      h <- list()
      cells.remove <- list()
      sc10x.temp <- list()
      thresh.l <- list()
251
      cutoff.l.count <- list()
252
      for (j in names(sc10x)){
253
254
255
256
257
258
259
        cutoff.l.count[[j]] <- NULL
        temp <- data.frame(sc10x[[j]][[i]])$nCount_RNA
        h[[i]] <- hist(temp,breaks=seq(min(temp),max(temp),length.out=6),plot=FALSE)
        cutoff.temp <- mean(c(h[[i]]$mids[which.max(h[[i]]$counts)],h[[i]]$mids[-which.max(h[[i]]$counts)][which.max(h[[i]]$counts[-which.max(h[[i]]$counts)])]))
        cells.remove[[j]] <- c(cells.remove[[j]],rownames(sc10x[[j]][["nCount_RNA"]])[sc10x[[j]][[i]][,1] > cutoff.temp])
        sc10x.temp[[j]] <- subset(sc10x[[j]],cells=setdiff(colnames(sc10x[[j]]),cells.remove[[j]]))
        thresh.l[[i]] <- scThresh(sc10x.temp,feature=i,sub="lower")
260
        cutoff.l.count[[j]] <- thresh.l[[i]][[j]]$threshold[thresh.l[[i]][[j]]$method=="MinErrorI"]
261
262
      }
    }
263
264
265
266
267
268
  }
  
  #Plot raw stats
  max.ct <- 0
  max.ft <- 0
  max.mt <- 0
269
  max.rb <- 0
270
271
272
273
274
275
276
277
278
279
  for (i in names(sc10x)){
    if (max.ct < max(sc10x[[i]][["nCount_RNA"]])){
      max.ct <- max(sc10x[[i]][["nCount_RNA"]])
    }
    if (max.ft < max(sc10x[[i]][["nFeature_RNA"]])){
      max.ft <- max(sc10x[[i]][["nFeature_RNA"]])
    }
    if (max.mt < max(sc10x[[i]][["percent.mito"]])){
      max.mt <- max(sc10x[[i]][["percent.mito"]])
    }
280
281
282
    if (max.rb < max(sc10x[[i]][["percent.ribo"]])){
      max.rb <- max(sc10x[[i]][["percent.ribo"]])
    }
283
284
285
286
  }
  max.ct <- max.ct*1.1
  max.ft <- max.ft*1.1
  max.mt <- max.mt*1.1
287
  max.rb <- max.rb*1.1
288
  cells.remove <- list()
289
  for (i in feature){
290
291
292
293
294
295
296
    max.y <- 0
    if (i == "nCount_RNA"){
      max.y <- max.ct
    } else if (i == "nFeature_RNA"){
      max.y <- max.ft
    } else if (i == "percent.mito"){
      max.y <- max.mt
297
298
    } else if (i == "percent.ribo"){
      max.y <- max.rb
299
300
301
302
303
304
305
    }
    plots.v <- list()
    densities.s <- list()
    plots.s <- list()
    for (j in names(sc10x)){
      sc10x.temp <- sc10x[[j]]
      plots.v[[j]] <- VlnPlot(object=sc10x.temp,features=i,pt.size=0.1,)+scale_x_discrete(labels=j)+scale_y_continuous(limits=c(0,max.y))+theme(legend.position="none",axis.text.x=element_text(hjust=0.5,angle=0))
306
      if (i %in% c("nFeature_RNA","percent.mito","percent.ribo","nCount_RNA")){
307
        if (i == "nFeature_RNA"){
Gervaise Henry's avatar
Gervaise Henry committed
308
          #cutoff.l <- thresh[[i]][[j]]$threshold[thresh[[i]][[j]]$method=="MinErrorI"]
309
          cutoff.h <- thresh[[i]][[j]]$threshold[thresh[[i]][[j]]$method=="RenyiEntropy"]
Gervaise Henry's avatar
Gervaise Henry committed
310
311
          cutoff.l <- 200
          #cutoff.h <- thresh[[i]][[j]]$threshold[thresh[[i]][[j]]$method=="Huang2"]
312
        } else if (i == "percent.mito") {
313
314
315
316
317
318
319
320
321
322
323
          h <- list()
          cells.remove <- list()
          sc10x.temps <- list()
          thresh.h <- list()
          temp <- data.frame(sc10x[[j]][[i]])$percent.mito
          h[[i]] <- hist(temp,breaks=100,plot=FALSE)
          cutoff.temp <- mean(c(h[[i]]$mids[which.max(h[[i]]$counts)],h[[i]]$mids[-which.max(h[[i]]$counts)][which.max(h[[i]]$counts[-which.max(h[[i]]$counts)])]))
          cells.remove[[j]] <- c(cells.remove[[j]],rownames(sc10x[[j]][["percent.mito"]])[sc10x[[j]][[i]][,1] < cutoff.temp])
          sc10x.temps[[j]] <- subset(sc10x[[j]],cells=setdiff(colnames(sc10x[[j]]),cells.remove[[j]]))
          thresh.h[[i]] <- scThresh(sc10x.temps,feature=i,sub="higher")
          cutoff.h <- thresh.h[[i]][[j]]$threshold[thresh.h[[i]][[j]]$method=="Triangle"]
Gervaise Henry's avatar
Gervaise Henry committed
324
          cutoff.l <- 0
325
326
327
328
        } else if (i == "percent.ribo") {
          cutoff.h <- thresh[[i]][[j]]$threshold[thresh[[i]][[j]]$method=="RenyiEntropy"]
          cutoff.l <- 0
        } else if (i == "nCount_RNA") {
329
330
          #cutoff.l <- thresh[[i]][[j]]$threshold[thresh[[i]][[j]]$method=="MinErrorI"]
          cutoff.l <- cutoff.l.count[[j]]
331
          cutoff.h <- max(sc10x[[j]][[i]])
332
333
        }
        plots.v[[j]] <- plots.v[[j]]+geom_hline(yintercept=cutoff.l,size=0.5,color="red")+geom_hline(yintercept=cutoff.h,size=0.5,color="red")
334
335
336
337
338
339
340
        if (i != "nCount_RNA"){
          densities.s[[j]] <- density(sc10x.temp$nCount_RNA,sc10x.temp[[i]][,1],n=1000)
          plots.s[[j]] <- ggplotGrob(ggplot(data.frame(cbind(sc10x.temp$nCount_RNA,sc10x.temp[[i]][,1])))+geom_point(aes(x=X1,y=X2,color=densities.s[[j]]),size=0.1)+scale_x_continuous(limits=c(0,max.ct))+scale_y_continuous(limits=c(0,max.y))+scale_color_viridis(option="inferno")+labs(x="nCount_RNA",y=i,color="Density")+ggtitle(j)+cowplot::theme_cowplot()+theme(plot.title=element_text(size=7.5),axis.title=element_text(size=7.5),axis.text=element_text(size=5,angle=45),legend.position="bottom",legend.title=element_text(size=5,face="bold",vjust=1),legend.text=element_text(size=5,angle=45))+guides(color=guide_colourbar(barwidth=5,barheight=0.5))+geom_hline(yintercept=cutoff.l,size=0.1,color="red")+geom_hline(yintercept=cutoff.h,size=0.1,color="red"))
        } else {
          densities.s[[j]] <- density(sc10x.temp$nFeature_RNA,sc10x.temp[[i]][,1],n=1000)
          plots.s[[j]] <- ggplotGrob(ggplot(data.frame(cbind(sc10x.temp$nFeature_RNA,sc10x.temp[[i]][,1])))+geom_point(aes(x=X1,y=X2,color=densities.s[[j]]),size=0.1)+scale_x_continuous(limits=c(0,max.ct))+scale_y_continuous(limits=c(0,max.y))+scale_color_viridis(option="inferno")+labs(x="nFeature_RNA",y=i,color="Density")+ggtitle(j)+cowplot::theme_cowplot()+theme(plot.title=element_text(size=7.5),axis.title=element_text(size=7.5),axis.text=element_text(size=5,angle=45),legend.position="bottom",legend.title=element_text(size=5,face="bold",vjust=1),legend.text=element_text(size=5,angle=45))+guides(color=guide_colourbar(barwidth=5,barheight=0.5))+geom_hline(yintercept=cutoff.l,size=0.1,color="red")+geom_hline(yintercept=cutoff.h,size=0.1,color="red"))
        }
341
        cells.remove[[j]] <- c(cells.remove[[j]],rownames(sc10x[[j]][[i]])[sc10x[[j]][[i]][,1] < cutoff.l | sc10x[[j]][[i]][,1] > cutoff.h])
342
      }
Gervaise Henry's avatar
Gervaise Henry committed
343
      ggsave(paste0("./analysis/qc/Violin_qc.raw.",i,".",j,".eps"),plot=plots.v[[j]])
344
      if (i %in% c("nFeature_RNA","percent.mito","percent.ribo","nCount_RNA")){
Gervaise Henry's avatar
Gervaise Henry committed
345
346
        ggsave(paste0("./analysis/qc/Plot_qc.raw.",i,".",j,".eps"),plot=plots.s[[j]])
      }
347
    }
348
  }
349
  
350
351
352
353
354
355
356
357
358
  #Record cell/gene counts pre and post filtering
  counts.cell.raw <- list()
  counts.gene.raw <- list()
  sc10x.sub <- list()
  counts.cell.filtered <- list()
  counts.gene.filtered <- list()
  for (i in names(sc10x)){
    counts.cell.raw[i] <- ncol(GetAssayData(object=sc10x[[i]],slot="counts"))
    counts.gene.raw[i] <- nrow(GetAssayData(object=sc10x[[i]],slot="counts"))
359
    sc10x.sub[[i]] <- subset(sc10x[[i]],cells=setdiff(colnames(sc10x[[i]]),cells.remove[[i]]))
360
361
362
    counts.cell.filtered[i] <- ncol(GetAssayData(object=sc10x.sub[[i]],slot="counts"))
    counts.gene.filtered[i] <- nrow(GetAssayData(object=sc10x.sub[[i]],slot="counts"))
  }
363
364
  
  #Plot filtered stats
365
366
367
  max.ct <- 0
  max.ft <- 0
  max.mt <- 0
368
  max.rb <- 0
369
370
371
372
373
374
375
376
377
378
  for (i in names(sc10x)){
    if (max.ct < max(sc10x.sub[[i]][["nCount_RNA"]])){
      max.ct <- max(sc10x.sub[[i]][["nCount_RNA"]])
    }
    if (max.ft < max(sc10x.sub[[i]][["nFeature_RNA"]])){
      max.ft <- max(sc10x.sub[[i]][["nFeature_RNA"]])
    }
    if (max.mt < max(sc10x.sub[[i]][["percent.mito"]])){
      max.mt <- max(sc10x.sub[[i]][["percent.mito"]])
    }
379
380
381
    if (max.rb < max(sc10x.sub[[i]][["percent.ribo"]])){
      max.rb <- max(sc10x.sub[[i]][["percent.ribo"]])
    }
382
  }
383
384
385
  max.ct <- max.ct*1.1
  max.ft <- max.ft*1.1
  max.mt <- max.mt*1.1
386
387
  max.rb <- max.rb*1.1
  for (i in feature){
388
389
390
391
392
393
394
    max.y <- 0
    if (i == "nCount_RNA"){
      max.y <- max.ct
    } else if (i == "nFeature_RNA"){
      max.y <- max.ft
    } else if (i == "percent.mito"){
      max.y <- max.mt
395
396
    } else if (i == "percent.ribo"){
      max.y <- max.rb
397
398
399
400
401
402
403
    }
    plots.v <- list()
    densities.s <- list()
    plots.s <- list()
    for (j in names(sc10x.sub)){
      sc10x.temp <- sc10x.sub[[j]]
      plots.v[[j]] <- VlnPlot(object=sc10x.temp,features=i,pt.size=0.1,)+scale_x_discrete(labels=j)+scale_y_continuous(limits=c(0,max.y))+theme(legend.position="none",axis.text.x=element_text(hjust=0.5,angle=0))
404
      if (i != "nCount_RNA"){
405
        densities.s[[j]] <- density(sc10x.temp$nCount_RNA,sc10x.temp[[i]][,1],n=1000)
406
407
408
409
        plots.s[[j]] <- ggplotGrob(ggplot(data.frame(cbind(sc10x.temp$nCount_RNA,sc10x.temp[[i]][,1])))+geom_point(aes(x=X1,y=X2,color=densities.s[[j]]),size=0.1)+scale_x_continuous(limits=c(0,max.ct))+scale_y_continuous(limits=c(0,max.y))+scale_color_viridis(option="inferno")+labs(x="nCount_RNA",y=i,color="Density")+ggtitle(j)+cowplot::theme_cowplot()+theme(plot.title=element_text(size=7.5),axis.title=element_text(size=7.5),axis.text=element_text(size=5,angle=45),legend.position="bottom",legend.title=element_text(size=5,face="bold",vjust=1),legend.text=element_text(size=5,angle=45))+guides(color=guide_colourbar(barwidth=5,barheight=0.5)))
      } else {
        densities.s[[j]] <- density(sc10x.temp$nFeature_RNA,sc10x.temp[[i]][,1],n=1000)
        plots.s[[j]] <- ggplotGrob(ggplot(data.frame(cbind(sc10x.temp$nFeature_RNA,sc10x.temp[[i]][,1])))+geom_point(aes(x=X1,y=X2,color=densities.s[[j]]),size=0.1)+scale_x_continuous(limits=c(0,max.ct))+scale_y_continuous(limits=c(0,max.y))+scale_color_viridis(option="inferno")+labs(x="nFeature_RNA",y=i,color="Density")+ggtitle(j)+cowplot::theme_cowplot()+theme(plot.title=element_text(size=7.5),axis.title=element_text(size=7.5),axis.text=element_text(size=5,angle=45),legend.position="bottom",legend.title=element_text(size=5,face="bold",vjust=1),legend.text=element_text(size=5,angle=45))+guides(color=guide_colourbar(barwidth=5,barheight=0.5)))
410
      }
411
      ggsave(paste0("./analysis/qc/Violin_qc.filtered.",i,".",j,".eps"),plot=plots.v[[j]])
412
      if (i %in% c("nFeature_RNA","percent.mito","percent.ribo","nCount_RNA")){
413
414
415
        ggsave(paste0("./analysis/qc/Plot_qc.filtered.",i,".",j,".eps"),plot=plots.s[[j]])
      }

416
417
418
    }
  }
  
419
420
421
422
423
424
425
426
427
428
429
  
  results <- list(
    sc10x=sc10x.sub,
    counts.cell.raw=counts.cell.raw,
    counts.gene.raw=counts.gene.raw,
    counts.cell.filtered=counts.cell.filtered,
    counts.gene.filtered=counts.gene.filtered
  )
  return(results)
}

430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
scThresh <- function(sc10x,feature,sub=FALSE){
  #Calculate thresholds and cutoffs
  
  #Inputs:
  #sc10x = Seruat object
  #feature = feature to threshold
  #sub = Subfolder to save output files
  
  #Outputs:
  #result = Threshold data
  
  
  #Make folders
  if (sub==FALSE){
    folder <- "./analysis/qc/cutoffs/"
  } else {
    folder <- paste0("./analysis/qc/cutoffs/",sub,"/")
    if (!dir.exists(folder)){
      dir.create(folder)
    }
450
451
  }
  
452
453
454
455
456
457
458
  #Calculate range of histogram based threholding and manually select methods for cutoffs
  scale <- list()
  scale.scaled <- list()
  h <- list()
  thresh <-list()
  cutoff.l <- list()
  cutoff.h <- list()
459
  thresh_methods <- c("IJDefault","Huang","Huang2","IsoData","Li","Mean","MinErrorI","Moments","Otsu","Percentile","RenyiEntropy","Shanbhag","Triangle")#,"Intermodes"
460
461
462
463
  for (i in names(sc10x)){
    scale[[i]] <- data.frame(Score=sc10x[[i]][[feature]])
    colnames(scale[[i]]) <- "Score"
    scale[[i]] <- data.frame(Score=scale[[i]]$Score[!is.na(scale[[i]]$Score)])
Gervaise Henry's avatar
Gervaise Henry committed
464
465
466
    scale.scaled[[i]] <- as.integer((scale[[i]]$Score-min(scale[[i]]$Score))/(max(scale[[i]]$Score)-min(scale[[i]]$Score))*360)
    #scale.scaled[[i]] <- as.integer(scales::rescale(scale[[i]]$Score,to=c(0,1))*360)
    h[[i]] <- hist(scale[[i]]$Score,breaks=100,plot=FALSE)
Gervaise Henry's avatar
Gervaise Henry committed
467
    thresh[[i]] <- purrr::map_chr(thresh_methods,~auto_thresh(scale.scaled[[i]],.)) %>% tibble(method = thresh_methods, threshold = .)
468
    thresh[[i]]$threshold <- as.numeric(thresh[[i]]$threshold)
Gervaise Henry's avatar
Gervaise Henry committed
469
470
    thresh[[i]]$threshold <- ((thresh[[i]]$threshold/360)*(max(scale[[i]]$Score)-min(scale[[i]]$Score)))+min(scale[[i]]$Score)
    #thresh[[i]] <- thresh[[i]] %>% mutate(threshold=(scales::rescale(as.numeric(threshold)/360,to=range(scale[[i]]$Score))))
471
472
473
474
475
476
477
478
479
    postscript(paste0(folder,"Hist_qc.",i,".",feature,".eps"))
    plot(h[[i]],main=paste0("Histogram of ",feature," of sample ",i),xlab=feature)
    abline(v=thresh[[i]]$threshold)
    mtext(thresh[[i]]$method,side=1,line=2,at=thresh[[i]]$threshold,cex=0.5)
    dev.off()
  }
  
  
  return(thresh)
480
}
481

482
scCellCycle <- function(sc10x,sub=FALSE,sp="hu"){
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
  #Runs Seurat based PCA analysis for cell cycle ID
  
  #Inputs:
  #sc10x = Seruat object
  #sub = Subfolder to save output files
  
  #Outputs:
  #results[1] = Seurat object
  #results[2] = s genes
  #results[3] = g2m genes
  
  #Make sub-folders if necessary
  if (sub==FALSE){
    folder <- "./analysis/qc/cellcycle/"
  } else {
    folder <- paste0("./analysis/qc/cellcycle/",sub,"/")
    if (!dir.exists(folder)){
      dir.create(folder)
    }}
  
  #score cell cycle
  genes.cc <- readLines(con="./genesets/regev_lab_cell_cycle_genes.txt")
  genes.s <- genes.cc[1:43]
  genes.g2m <- genes.cc[44:97]
  sc10x <- NormalizeData(object=sc10x,verbose=FALSE)
  sc10x <- ScaleData(object=sc10x,do.par=TRUE,num.cores=45,verbose=FALSE)
  sc10x <- CellCycleScoring(object=sc10x,s.features=genes.s,g2m.features=genes.g2m,set.ident=TRUE)
  
  #plot cell cycle specific genes
512
513
514
515
516
517
518
  if (sp=="hu"){
    genes=c("PCNA","TOP2A","MCM6","MKI67")
    postscript(paste0(folder,"Violin_cc.Raw.eps"))
    plot <- VlnPlot(object=sc10x,features=genes,ncol=2,pt.size=1)
    plot(plot)
    dev.off()
  }
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
  
  # sc10x <- RunPCA(object=sc10x,features=c(genes.s,genes.g2m),npcs=2,verbose=FALSE)
  # postscript(paste0(folder,"PCA_cc.Raw.eps"))
  # plot <- DimPlot(object=sc10x,reduction="pca")
  # plot(plot)
  # dev.off()
  # gc()
  # sc10x <- ScaleData(object=sc10x,vars.to.regress=c("S.Score","G2M.Score"),do.par=TRUE,num.cores=45,verbose=TRUE)
  # gc()
  # sc10x <- RunPCA(object=sc10x,features=c(genes.s,genes.g2m),npcs=2,verbose=FALSE)
  # postscript(paste0(folder,"PCA_cc.Norm.eps"))
  # plot <- DimPlot(object=sc10x,reduction="pca")
  # plot(plot)
  # dev.off()
  
  results <- list(
    sc10x=sc10x,
    genes.s=genes.s,
    genes.g2m=genes.g2m
  )
  return(results)
}


543
scPC <- function(sc10x,pc=50,hpc=0.9,file="pre.stress",print="tsne"){
544
545
546
547
548
549
550
551
552
553
554
555
556
557
  #Scale Seurat object & calculate # of PCs to use
  
  #Inputs:
  #sc10x = Seruat object
  #pc = number of PCs to cacluate
  #hpc = max variance cutoff for PCs to use"
  #file = file for output
  
  #Outputs:
  #result[1] = Seurat object
  #result[2] = # of PCs to use
  
  #Run PCA
  Idents(object=sc10x) <- "ALL"
558
  sc10x <- RunPCA(object=sc10x,npcs=pc,verbose=FALSE,assay="integrated")
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
  
  #Calculate PCs to use
  pc.use <- sc10x[["pca"]]@stdev^2
  pc.use <- pc.use/sum(pc.use)
  pc.use <- cumsum(pc.use)
  pc.use <- min(which(pc.use>=hpc))
  
  postscript(paste0("./analysis/qc/Plot_PCElbow_",file,".eps"))
  plot <- ElbowPlot(object=sc10x,ndims=pc)
  plot <- plot+geom_vline(xintercept=pc.use,size=1,color="red")
  plot(plot)
  dev.off()
  
  results <- list(
    sc10x=sc10x,
    pc.use=pc.use
  )
  return(results)
}


580
581
scCCA <-  function(sc10x.l){
  for (i in 1:length(sc10x.l)){
582
    #sc10x.l[[i]] <- NormalizeData(sc10x.l[[i]],verbose=FALSE)
583
    gc()
584
585
    #sc10x.l[[i]] <- ScaleData(sc10x.l[[i]],vars.to.regress=c("nFeature_RNA","percent.mito"),verbose = FALSE)
    sc10x.l[[i]] <- SCTransform(sc10x.l[[i]],vars.to.regress=c("nFeature_RNA","percent.mito"),verbose=FALSE,assay="RNA")
586
    gc()
587
    #sc10x.l[[i]] <- FindVariableFeatures(sc10x.l[[i]],selection.method="vst",nfeatures=2000,verbose=FALSE)
588
589
  }
  
590
591
  sc10x.features <- SelectIntegrationFeatures(object.list=sc10x.l,nfeatures=3000)
  sc10x.l <- PrepSCTIntegration(object.list=sc10x.l,anchor.features=sc10x.features,verbose=FALSE)
592
593
594
595

  sc10x.l <- lapply(sc10x.l,FUN=function(x) { RunPCA(x,features=sc10x.features,verbose=FALSE) })
  
  sc10x.anchors <- FindIntegrationAnchors(object.list=sc10x.l,normalization.method="SCT",anchor.features=sc10x.features,verbose=FALSE,reduction="rpca",dims=1:30)
596
  sc10x <- IntegrateData(anchorset=sc10x.anchors,normalization.method="SCT",verbose=FALSE)
597
  
598
599
600
601
602
603
  #sc10x <- FindIntegrationAnchors(object.list=sc10x.l,dims=1:30,scale=FALSE)
  #sc10x <- IntegrateData(anchorset=sc10x,dims=1:30)
  
  #gc()
  #sc10x <- ScaleData(object=sc10x,do.par=TRUE,num.cores=45,verbose=FALSE,assay="integrated")
  #gc()
604
  
605
  gc()
606
  sc10x <- SCTransform(sc10x,vars.to.regress=c("nFeature_RNA","percent.mito"),verbose=FALSE,return.only.var.genes=FALSE,assay="RNA")
607
  gc()
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
  
  return(sc10x)
}


scCluster <- function(sc10x,res=0.1,red="pca",dim,print="tsne",folder=FALSE){
  #Cluster Seurat object and produce visualizations
  
  #Inputs:
  #sc10x = Seruat object
  #res = resolution to calculate clustering
  #red = rediction type to use for clustering
  #dim = number of dimentions to use for display
  #print = dimentionality reduction to use for display
  #folder = folder for output
  
  #Outputs:
  #result = Seurat object
  
  #Create subfolder if necessary
  if (folder==FALSE){
    sub <- ""
  } else {
    if (!dir.exists(paste0("./analysis/vis/",folder))){
      dir.create(paste0("./analysis/vis/",folder))
    }
    sub <- paste0(folder,"/")
    
  }
  
638
639
  DefaultAssay(sc10x) <- "integrated"

640
  #Calculste Vis
641
642
  sc10x <- RunTSNE(sc10x,dims=1:dim,reduction="pca",assay="integrated")
  sc10x <- RunUMAP(sc10x,dims=1:dim,reduction="pca",assay="integrated")
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
  
  sc10x <- FindNeighbors(sc10x,reduction=red,verbose=FALSE)
  
  for (i in res){
    sc10x <- FindClusters(sc10x,resolution=i,verbose=FALSE)
    
    plot1 <- DimPlot(sc10x,reduction="pca",label=TRUE,repel=TRUE)+theme(legend.position="none")
    plot2 <- DimPlot(sc10x,reduction="tsne",label=TRUE,repel=TRUE)+theme(legend.position="none")
    plot3 <- DimPlot(sc10x,reduction="umap",label=TRUE,repel=TRUE)+theme(legend.position="none")
    
    if (print=="tsne"){
      postscript(paste0("./analysis/vis/",sub,"tSNE_",i,".eps"))
      print(print2)
      dev.off()
    } else if (print=="umap"){
      postscript(paste0("./analysis/vis/",sub,"UMAP_",i,".eps"))
      print(print3)
      dev.off()
    } else if (print=="2"){
      plot2 <- plot2+theme(legend.position="none")
      plot3 <- plot3+theme(legend.position="none")
      postscript(paste0("./analysis/vis/",sub,"2Vis_",i,".eps"))
      grid.arrange(plot2,plot3,ncol=1)
      dev.off()
    } else if (print=="3"){
      plot1 <- plot1+theme(legend.position="none")
      plot2 <- plot2+theme(legend.position="none")
      plot3 <- plot3+theme(legend.position="none")
      postscript(paste0("./analysis/vis/",sub,"3Vis_",i,".eps"))
      grid.arrange(plot1,plot2,plot3,ncol=1)
      dev.off()
    }}
  
Gervaise Henry's avatar
Gervaise Henry committed
676
  for (i in c("samples","HTO_maxID","hash.ID")[c("samples","HTO_maxID","hash.ID") %in% colnames(sc10x@meta.data)]){
Gervaise Henry's avatar
Gervaise Henry committed
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
    plot1 <- DimPlot(sc10x,reduction="pca",group.by=i)
    plot2 <- DimPlot(sc10x,reduction="tsne",group.by=i)
    plot3 <- DimPlot(sc10x,reduction="umap",group.by=i)
    legend <- cowplot::get_legend(plot1)
    
    if (print=="tsne"){
      postscript(paste0("./analysis/vis/",sub,"tSNE_",i,".eps"))
      grid.arrange(plot2,legend,ncol=1)
      dev.off()
    } else if (print=="umap"){
      postscript(paste0("./analysis/vis/",sub,"UMAP_",i,".eps"))
      grid.arrange(plot3,legend,ncol=1)
      dev.off()
    } else if (print=="2"){
      plot2 <- plot2+theme(legend.position="none")
      plot3 <- plot3+theme(legend.position="none")
      postscript(paste0("./analysis/vis/",sub,"2Vis_",i,".eps"))
      grid.arrange(plot2,plot3,legend,ncol=1)
      dev.off()
    } else if (print=="3"){
      plot1 <- plot1+theme(legend.position="none")
      plot2 <- plot2+theme(legend.position="none")
      plot3 <- plot3+theme(legend.position="none")
      postscript(paste0("./analysis/vis/",sub,"3Vis_",i,".eps"))
      grid.arrange(plot1,plot2,plot3,legend,ncol=1)
      dev.off()
    }
704
  }
Gervaise Henry's avatar
Gervaise Henry committed
705
  
706
  DefaultAssay(sc10x) <- "SCT"
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
  
  return(sc10x)
}


scScore <- function(sc10x,score,geneset,cut.pt=0.9,anchor=FALSE){
  #Runs custom PCA analysis for stress ID
  
  #Inputs:
  #sc10x = Seruat object
  #score = name of geneset for scoring
  #geneset = geneset to use for ID
  #cut.pt = % of cells to keep
  
  #Outputs:
  #results[1] = Seurat object (original + score)
  #results[2] = Seurat object (negatively filtered)
  #results[3] = Seurat object (positively filtered)
  
  #Make subdirectory
  if (!dir.exists(paste0("./analysis/score_id/",score))){
    dir.create(paste0("./analysis/score_id/",score))
  }
  if (!dir.exists(paste0("./analysis/vis/",score))){
    dir.create(paste0("./analysis/vis/",score))
  }
  
  #Score geneset
735
  sc10x <- AddModuleScore(object=sc10x,features=geneset,name=score,assay="SCT")
736
737
738
739
  Idents(object=sc10x) <- paste0(score,"1")
  
  #CDF
  cdf <- ecdf(as.numeric(levels(sc10x)))
Gervaise Henry's avatar
Gervaise Henry committed
740
  if (cut.pt == "renyi"){
741
742
743
744
745
746

        h <- hist(data.frame(sc10x[[paste0(score,"1")]])[,paste0(score,"1")],breaks=1000,plot=FALSE)
	cutoff.temp <- mean(c(h$mids[which.max(h$counts)],h$mids[-which.max(h$counts)][which.max(h$counts[-which.max(h$counts)])]))
        cells.remove <- rownames(sc10x[[paste0(score,"1")]])[sc10x[[paste0(score,"1")]][,1] < cutoff.temp]
        sc10x.temp <- subset(sc10x,cells=setdiff(colnames(sc10x),cells.remove))
 
747
    thresh <- list()
748
    thresh[["all"]] <- scThresh(list(all=sc10x.temp),feature=paste0(score,"1"),sub=score)
Gervaise Henry's avatar
Gervaise Henry committed
749
    cut.x <- thresh$all$all$threshold[thresh$all$all$method=="RenyiEntropy"]
750
751
752
753
  } else {
    cut.x <- quantile(cdf,probs=cut.pt)
    cut.x <- unname(cut.x)
  }
754
755
756
757
758
759
760
761
762
763
764
765
766
767
  postscript(paste0("./analysis/score_id/",score,"/CDF_",score,".eps"))
  plot(cdf,main=paste0("Cumulative Distribution of ",score," Score"),xlab=paste0(score," Score"),ylab="CDF")
  abline(v=cut.x,col="red")
  dev.off()  
  
  #KDE
  postscript(paste0("./analysis/score_id/",score,"/Histo_",score,".eps"))
  plot(ggplot(data.frame(Score=as.numeric(levels(sc10x))),aes(x=Score))+geom_histogram(bins=100,aes(y=..density..))+geom_density()+geom_vline(xintercept=cut.x,size=1,color="red")+ggtitle(paste0(score," Score"))+cowplot::theme_cowplot())
  dev.off()
  
  Idents(object=sc10x) <- "ALL"
  predicate <- paste0(score,"1 >= ",cut.x)
  Idents(object=sc10x, cells = WhichCells(object=sc10x,expression= predicate)) <- score
  sc10x[[score]] <- Idents(object=sc10x)
768
  Idents(sc10x) <- score
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
  sc10x.negative <- subset(x=sc10x,idents="ALL")
  sc10x.positive <- subset(x=sc10x,idents=score)
  
  #Generate plots
  postscript(paste0("./analysis/vis/",score,"/3Vis_",score,".eps"))
  plot1 <- DimPlot(sc10x,reduction="pca",label=TRUE,repel=TRUE)+theme(legend.position="none")+ggtitle("ALL")
  plot2 <- DimPlot(sc10x.negative,reduction="pca",label=TRUE,repel=TRUE)+theme(legend.position="none")
  plot3 <- DimPlot(sc10x.positive,reduction="pca",label=TRUE,repel=TRUE)+theme(legend.position="none")
  plot4 <- DimPlot(sc10x,reduction="tsne",label=TRUE,repel=TRUE)+theme(legend.position="none")+ggtitle("Negative")
  plot5 <- DimPlot(sc10x.negative,reduction="tsne",label=TRUE,repel=TRUE)+theme(legend.position="none")
  plot6 <- DimPlot(sc10x.positive,reduction="tsne",label=TRUE,repel=TRUE)+theme(legend.position="none")
  plot7 <- DimPlot(sc10x,reduction="umap",label=TRUE,repel=TRUE)+theme(legend.position="none")+ggtitle("Positive")
  plot8 <- DimPlot(sc10x.negative,reduction="umap",label=TRUE,repel=TRUE)+theme(legend.position="none")
  plot9 <- DimPlot(sc10x.positive,reduction="umap",label=TRUE,repel=TRUE)+theme(legend.position="none")
  grid.arrange(plot1,plot2,plot3,plot4,plot5,plot6,plot7,plot8,plot9,ncol=3)
  dev.off()
  
  #Generate violin plot of gene exvpression
  if (anchor!=FALSE){
    postscript(paste0("./analysis/score_id/",score,"/Violin_",score,".eps"))
789
    plot <- VlnPlot(object=sc10x,features=anchor,pt.size=0.1,assay="SCT")
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
    plot(plot)
    dev.off()
  }
  
  results <- list(
    sc10x <- sc10x,
    sc10x.negative <- sc10x.negative,
    sc10x.positive <- sc10x.positive
  )
  return(results)
}


scQuSAGE <- function(sc10x,gs,save=FALSE,type,id,ds=0,nm="pops",print="tsne"){
  #Runs QuSAGE
  
  #Inputs:
  #sc10x = Seruat object
  #gs = geneset to use for correlation
  #save = save ID
  #type = type of qusage to run (id: create ID's based on run, sm: cor only using small genesets, lg: cor only with large genesets)
  #id = ident to use
  #nm = name of test
  #print = dimentionality reduction to use for display
  
  #Outputs:
  #results[1] = Seurat object
  #results[2] = correlation table
  #results[3] = correlation results
  
  if (!dir.exists(paste0("./analysis/cor/",nm))){
    dir.create(paste0("./analysis/cor/",nm))
  }
  if (!dir.exists(paste0("./analysis/cor/",nm,"/geneset"))){
    dir.create(paste0("./analysis/cor/",nm,"/geneset"))
  }
  if (!dir.exists(paste0("./analysis/cor/",nm,"/cluster"))){
    dir.create(paste0("./analysis/cor/",nm,"/cluster"))
  }
  if (!dir.exists(paste0("./analysis/vis/",nm))){
    dir.create(paste0("./analysis/vis/",nm))
  }
  
  Idents(object=sc10x) <- id
  number.clusters <- length(unique(levels(x=sc10x)))
  
  labels <- paste0("Cluster_",as.vector(Idents(object=sc10x)))
  
  cell.sample <- NULL
  for (i in unique(labels)){
    cell <- WhichCells(sc10x,ident=sub("Cluster_","",i))
    if (length(cell)>ds & ds!=0){
842
      set.seed(71682)
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
      rnd <- sample(1:length(cell),ds)
      cell <- cell[rnd]
    }
    cell.sample <- c(cell.sample,cell)
  }
  data <- as.data.frame(as.matrix(GetAssayData(sc10x[,colnames(sc10x) %in% cell.sample])))
  labels <- labels[colnames(sc10x) %in% cell.sample]
  groups <- sort(unique(labels))
  
  col <- hcl(h=(seq(15,375-375/length(groups),length=length(groups))),c=100,l=65)
  
  #Make labels for QuSAGE
  clust <- list()
  clust.comp <- list()
  for (i in groups){
    t <- labels
    t[t != i] <- "REST"
    clust[i] <- list(i=t)
    rm(t)
    clust.comp[i] <- paste0(i,"-REST")
  }
  
  #Run QuSAGE
  for (i in groups){
    assign(paste0("results.",i),qusage(data,unlist(clust[i]),unlist(clust.comp[i]),gs))
  }
  
  #Generate ID table
  results.cor <- NULL
  results.cor <- qsTable(get(paste0("results.",groups[1])),number=length(gs))
  results.cor$Cluster <- groups[1]
  for (i in groups[-1]){
    qs <- qsTable(get(paste0("results.",i)),number=length(gs))
    qs$Cluster <- i
    results.cor <- rbind(results.cor,qs)
  }
  results.cor <- results.cor[,-3]
  rownames(results.cor) <- NULL
  
  results.clust.id <- NULL
883
884
885
886
  #if (max(results.cor[results.cor[,4]==groups[1] & results.cor[,3]<=0.05,][,2],na.rm=TRUE)>=0){
  #  results.clust.id <- results.cor[results.cor[,4]==groups[1] & results.cor[,3]<=0.05,][which.max(results.cor[results.cor[,4]==groups[1] & results.cor[,3]<=0.05,][,2]),]
  if (max(results.cor[results.cor[,4]==groups[1],][,2],na.rm=TRUE)>=0){
    results.clust.id <- results.cor[results.cor[,4]==groups[1],][which.max(results.cor[results.cor[,4]==groups[1],][,2]),]
887
888
889
890
891
892
893
894
  } else {
    results.clust.id$pathway.name <- "Unknown"
    results.clust.id$log.fold.change <- 0
    results.clust.id$FDR <- 0
    results.clust.id$Cluster <- groups[1]
    results.clust.id <- as.data.frame(results.clust.id)
  }
  for (i in groups[-1]){
895
896
897
898
    #if (max(results.cor[results.cor[,4]==i & results.cor[,3]<=0.05,][,2],na.rm=TRUE)>=0){
    #  results.clust.id <- rbind(results.clust.id,results.cor[results.cor[,4]==i & results.cor[,3]<=0.05,][which.max(results.cor[results.cor[,4]==i & results.cor[,3]<=0.05,][,2]),])
    if (max(results.cor[results.cor[,4]==i,][,2],na.rm=TRUE)>=0){
      results.clust.id <- rbind(results.clust.id,results.cor[results.cor[,4]==i,][which.max(results.cor[results.cor[,4]==i,][,2]),])
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
    } else {
      results.clust.id <- rbind(results.clust.id,data.frame(pathway.name="Unknown",log.fold.change=0,FDR=0,Cluster=i))
    }}
  rownames(results.clust.id) <- NULL
  
  #Determine axes for correlation plots
  max.x.rg <- 0
  min.x.rg <- 0
  max.y.rg <- 0
  for (i in groups){
    qs <- get(paste0("results.",i))
    if (max(qs$path.mean)>max.x.rg){
      max.x.rg <- max(qs$path.mean)
    }
    if (min(qs$path.mean)<min.x.rg){
      min.x.rg <- min(qs$path.mean)
    }
    if (max(qs$path.PDF)>max.y.rg){
      max.y.rg <- max(qs$path.PDF)
    }}
  if (type=="sm"){
920
921
922
923
924
925
926
927
928
929
930
931
932
933
    #Plot correlation plots by geneset
    for (i in 1:length(gs)){
      postscript(paste0("./analysis/cor/",nm,"/geneset/QuSAGE_",nm,".",names(gs)[i],".eps"))
      for (j in groups){
        qs <- get(paste0("results.",j))
        if (j==groups[1]){
          plotDensityCurves(qs,path.index=i,col=col[match(j,groups)],main=names(gs)[i],xlim=c(min.x.rg-0.05,max.x.rg+0.05),ylim=c(0,50*ceiling(max.y.rg/50)),xlab="Gene Set Activation",lwd=5,cex.main=2.5,cex.axis=1.5,cex.lab=2)
        } else {
          plotDensityCurves(qs,path.index=i,add=TRUE,col=col[match(j,groups)],lwd=5)
        }}
      legend("topright",col=col,legend=groups,lty=1,lwd=5,cex=2,ncol=1,bty="n",pt.cex=2)
      box(lwd=5)
      dev.off()
    }
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
    #Plot correlation plots by cluster
    for (i in groups){
      qs <- get(paste0("results.",i))
      postscript(paste0("./analysis/cor/",nm,"/cluster/QuSAGE_",nm,"_",i,".eps"))
      for (j in 1:length(gs)){
        if (j==1){
          plotDensityCurves(qs,path.index=j,col=viridis(length(gs))[j],main=i,xlim=c(min.x.rg-0.05,max.x.rg+0.05),ylim=c(0,50*ceiling(max.y.rg/50)),xlab="Gene Set Activation",lwd=5,cex.main=2.5,cex.axis=1.5,cex.lab=2)
        } else {
          plotDensityCurves(qs,path.index=j,add=TRUE,col=viridis(length(gs))[j],lwd=5)
        }}
      legend("topright",col=viridis(length(gs)),legend=names(gs),lty=1,lwd=5,cex=1,ncol=2,bty="n",pt.cex=2)
      box(lwd=5)
      dev.off()
    }} else {
      for (i in groups){
        qs <- get(paste0("results.",i))
        postscript(paste0("./analysis/cor/",nm,"/cluster/QuSAGE_",nm,"_",i,".eps"))
        plotCIs(qs,path.index=1:numPathways(qs),cex.lab=1.5)
        dev.off()
      }}
  
  if (save==TRUE){
    merge.cluster <- NULL
    for (i in groups){
958
959
960
961
      #if (max(qsTable(get(paste0("results.",i)),number=length(gs))[qsTable(get(paste0("results.",i)),number=length(gs))[,4]<=0.05,][,2],na.rm=TRUE)>=0){
      #  sc10x<-eval(parse(text=paste0("RenameIdents(object=sc10x,'",sub("Cluster_","",i),"' = '",qsTable(get(paste0("results.",i)),number=length(gs))[qsTable(get(paste0("results.",i)),number=length(gs))[2]==max(qsTable(get(paste0("results.",i)),number=length(gs))[qsTable(get(paste0("results.",i)),number=length(gs))[,4]<=0.05,][,2],na.rm=TRUE)][1],"')")))
      if (max(qsTable(get(paste0("results.",i)),number=length(gs))[,2],na.rm=TRUE)>=0){
        sc10x<-eval(parse(text=paste0("RenameIdents(object=sc10x,'",sub("Cluster_","",i),"' = '",qsTable(get(paste0("results.",i)),number=length(gs))[qsTable(get(paste0("results.",i)),number=length(gs))[2]==max(qsTable(get(paste0("results.",i)),number=length(gs))[,2],na.rm=TRUE)][1],"')")))
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
      } else {
        sc10x<-eval(parse(text=paste0("RenameIdents(object=sc10x,'",sub("Cluster_","",i),"' = 'Unknown')")))
      }}
    sc10x[[nm]] <- Idents(object=sc10x)
  }
  
  plot1 <- DimPlot(sc10x,reduction="pca",label=TRUE,repel=TRUE)+theme(legend.position="none")
  plot2 <- DimPlot(sc10x,reduction="tsne",label=TRUE,repel=TRUE)+theme(legend.position="none")
  plot3 <- DimPlot(sc10x,reduction="umap",label=TRUE,repel=TRUE)+theme(legend.position="none")
  if (print=="tsne"){
    postscript(paste0("./analysis/vis/",nm,"/tSNE_",id,"_",nm,".eps"))
    print(print2)
    dev.off()
  } else if (print=="umap"){
    postscript(paste0("./analysis/vis/",nm,"/UMAP_",id,"_",nm,".eps"))
    print(print3)
    dev.off()
  } else if (print=="2"){
    postscript(paste0("./analysis/vis/",nm,"/2Vis_",id,"_",nm,".eps"))
    grid.arrange(plot2,plot3,ncol=1)
    dev.off()
  } else if (print=="3"){
    postscript(paste0("./analysis/vis/",nm,"/3Vis_",id,"_",nm,".eps"))
    grid.arrange(plot1,plot2,plot3,ncol=1)
    dev.off()
  }
  
  results <- list(
    sc10x=sc10x,
    results.cor=results.cor,
    results.clust.id=results.clust.id
  )
  names(results)=c("sc10x",paste0("results.cor.",nm),paste0("results.clust.",nm,".id"))
  return(results)
}
997

998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
scShinyOutput <- function(sc10x,anal="raw"){
  write_delim(as.data.frame(colnames(sc10x)),path=paste0("./analysis/shiny/",anal,"/outs/filtered_feature_bc_matrix/barcodes.tsv.gz"),delim="\t",col_names=FALSE)
  features <- rownames(sc10x)
  features <- c(features,c("nFeature","nCount","percent.mito","percent.ribo","Stress.score"))
  features <- data.frame(ENSG=features,Feature=features,Label="feature")
  write_delim(features,path=paste0("./analysis/shiny/",anal,"/outs/filtered_feature_bc_matrix/features.tsv.gz"),delim="\t",col_names=FALSE)
  exp <- GetAssayData(sc10x,slot="scale.data")
  exp.extra <- matrix(nrow=5,ncol=ncol(sc10x))
  exp.extra[1,] <- as.numeric(sc10x$nFeature_RNA)
  exp.extra[2,] <- as.numeric(sc10x$nCount_RNA)
  exp.extra[3,] <- as.numeric(sc10x$percent.mito)
  exp.extra[4,] <- as.numeric(sc10x$percent.ribo)
  exp.extra[5,] <- as.numeric(sc10x$Stress1)
  exp <- rbind(exp,exp.extra)
Gervaise Henry's avatar
Gervaise Henry committed
1012
  Matrix::writeMM(as(exp,"dgCMatrix"),file=paste0("./analysis/shiny/",anal,"/outs/filtered_feature_bc_matrix/matrix.mtx.gz"))
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
  for (i in c("pca","tsne","umap")){
    dr <- Embeddings(sc10x,i)
    if (i != "pca"){
      colnames(dr) <- c(paste0(toupper(i),"-",1:2))
    } else {
      dr <- dr[,1:10]
      colnames(dr) <- c(paste0(toupper(i),"-",1:10))
    }
    dr <- cbind(dr,Barcode=rownames(dr))
    dr <- dr[,c(3,1,2)]
    dr <- as.data.frame(dr,row.names=FALSE)
    if (i != "pca"){
1025
      write_csv(dr,paste0("./analysis/shiny/",anal,"/outs/analysis/",i,"/2_components/projection.csv"),col_names=TRUE)
1026
    } else {
1027
      write_csv(dr,paste0("./analysis/shiny/",anal,"/outs/analysis/",i,"/10_components/projection.csv"),col_names=TRUE)
1028
1029
    }
  }
1030
  sc10x <- NormalizeData(sc10x,assay="RNA")
Gervaise Henry's avatar
Gervaise Henry committed
1031
  clusters <- c("samples","samples_HTO",paste0("integrated_snn_res.",res),"lin","pops","leu","scDWSpr","HTO_maxID","hash.ID")
1032
1033
  clusters <- intersect(clusters,names(sc10x@meta.data))
  for (i in clusters){
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
    if (nrow(unique(sc10x[[i]]))>1){
      if (!dir.exists(paste0("./analysis/shiny/",anal,"/outs/analysis/clustering/",gsub("integrated_snn_res.","res_",i)))){
        dir.create(paste0("./analysis/shiny/",anal,"/outs/analysis/clustering/",gsub("integrated_snn_res.","res_",i)))
      }
      clust <- as.matrix(sc10x[[i]])
      colnames(clust) <- "Cluster"
      clust <- cbind(clust,Barcode=rownames(clust))
      clust <- clust[,c(2,1)]
      clust <- as.data.frame(clust,row.names=FALSE)
      clust[,2] <- paste0("Cluster ",clust[,2])
      write_csv(clust,paste0("./analysis/shiny/",anal,"/outs/analysis/clustering/",gsub("integrated_snn_res.","res_",i),"/clusters.csv"),col_names=TRUE)
  
      if (!dir.exists(paste0("./analysis/shiny/",anal,"/outs/analysis/diffexp/",gsub("integrated_snn_res.","res_",i)))){
        dir.create(paste0("./analysis/shiny/",anal,"/outs/analysis/diffexp/",gsub("integrated_snn_res.","res_",i)))
      }
      Idents(sc10x) <- i
      deg <- FindAllMarkers(sc10x,assay="RNA",slot="data",logfc.threshold=0,test.use="MAST",min.pct=0.25,min.diff.pct=0.25,max.cells.per.ident=500)
      dexp <- data.frame("Feature ID"=unique(deg$gene),"Feature Name"=unique(deg$gene))
      for (cluster in unique(deg$cluster)){
        dexp[,paste0("Cluster.",cluster,".Mean.Counts")] <- deg$pct.1[deg$cluster==cluster][match(dexp$Feature.ID,deg$gene[deg$cluster==cluster])]
        dexp[,paste0("Cluster.",cluster,".Log2.fold.change")] <- deg$avg_logFC[deg$cluster==cluster][match(dexp$Feature.ID,deg$gene[deg$cluster==cluster])]
        dexp[,paste0("Cluster.",cluster,".Adjusted.p.value")] <- deg$p_val_adj[deg$cluster==cluster][match(dexp$Feature.ID,deg$gene[deg$cluster==cluster])]
      }
      colnames(dexp) <- gsub("\\."," ",colnames(dexp))
      write_csv(dexp,paste0("./analysis/shiny/",anal,"/outs/analysis/diffexp/",gsub("integrated_snn_res.","res_",i),"/differential_expression.csv"),col_names=TRUE)
1059
    }
1060
  }
1061
}