QC-by-cell

This is the QC of cell/library by looking at the number of cell captured in each well, the ERCC ratio, the number of unmapped reads, and the number of mapped reads. Higher ERCC ratio indicates that the RNA amount from the cell is lower, which migh be caused by cell death. The higher number of unmapped reads indicates lower quality of library preparation.

Library QC

##### list all the files
 wanted.files <- list.files(path="/mnt/gluster/home/ptung/test/ERCC/", pattern= "[0-9].counts.txt.summary")

##### get info you need
 all.tables <- lapply(wanted.files,function(x){
     read.table(paste("/mnt/gluster/home/ptung/test/ERCC/",x,sep=""),header=TRUE)[c(1,4,5),]
 })

#### create ERCC ratios
ERCC.ratios <- lapply(all.tables,function(x){
    x[1,2]/(sum(x[,2]))
})
ERCC.ratios <- unlist(ERCC.ratios)

#### create unmapped ratios
unmapped.ratios <- lapply(all.tables,function(x){
    x[3,2]/(sum(x[,2]))
})
unmapped.ratios <- unlist(unmapped.ratios)

##### create number of mapped reads
num.of.mapped.reads <- lapply(all.tables,function(x){
    (sum(x[,2]))-x[3,2]
})
num.of.mapped.reads <- unlist(num.of.mapped.reads)

##### create the table with all into
data1 <- data.frame(wanted.files,ERCC.ratios,unmapped.ratios,num.of.mapped.reads)

##### remove 4 individual cells
data1 <- data1[(data1$num.of.mapped.reads < 50000000),]

##### cell number capture on C1
data.cell.num <- read.csv("/mnt/gluster/home/ptung/LCL/cell_number.csv")
data1 <- data.frame(data1,data.cell.num)

##### QC cell number
index.c <- data1$cell.num == 1

##### QC mapped reads
index1 <- (data1[,"num.of.mapped.reads"] > 2500000)
sum(index1)

[1] 81

##### QC ERCC ratios
index2 <- (data1[,"ERCC.ratios"] < 0.05)
sum(index2)

[1] 83

##### QC unmapped ratios
index3 <- (data1[,"unmapped.ratios"] < 0.3)
sum(index3)

[1] 83

##### QC final ( | or, & and)
index.final <- index1 & index2 & index3 & index.c
index.final

 [1]  TRUE  TRUE FALSE FALSE  TRUE  TRUE  TRUE FALSE  TRUE  TRUE  TRUE
[12]  TRUE  TRUE  TRUE  TRUE  TRUE FALSE  TRUE FALSE FALSE  TRUE  TRUE
[23]  TRUE  TRUE FALSE  TRUE  TRUE FALSE  TRUE  TRUE  TRUE  TRUE  TRUE
[34]  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE FALSE  TRUE  TRUE  TRUE  TRUE
[45]  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE
[56]  TRUE  TRUE FALSE  TRUE  TRUE  TRUE  TRUE FALSE  TRUE  TRUE FALSE
[67]  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE FALSE FALSE  TRUE  TRUE
[78]  TRUE  TRUE  TRUE FALSE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE
[89] FALSE  TRUE FALSE  TRUE  TRUE FALSE  TRUE FALSE

sum(index.final)

[1] 77

#######  keep the names of good cells
keep.cell <- data1$wanted.files[index.final]

######## plot cells with color coded
color.index <- as.numeric(index.final)+1

plot(data1[,c(4,2)],col=color.index)

plot of chunk unnamed-chunk-1

plot(data1[,c(4,3)],col=color.index)

plot of chunk unnamed-chunk-1

plot(data1[,4]/1000000,data1[,2]*100,col=color.index,pch=as.character(data1$cell.num),xlab="Number of mapped reads (million)",ylab="Spike-In (%)")
legend(12,25,c("Cells Removed","Cells Retained"),col=1:2,pch=20)

plot of chunk unnamed-chunk-1

plot(data1[,4]/1000000,data1[,3]*100,col=color.index,pch=as.character(data1$cell.num),xlab="Number of mapped reads (million)",ylab="Unmapped reads (%)")
legend(12,70,c("Cells Removed","Cells Retained"),col=1:2,pch=20)

plot of chunk unnamed-chunk-1

##### plot the good cells only
plot(data1[index.final,c(2,4)],pch=data1$cell.num[index.final],col=color.index[index.final])

plot of chunk unnamed-chunk-1

QC-by-cell

PoYuan Tung

2015-02-13

Library QC

Session information