CODEWITHSUNDEEP
rdata-visualizationcorrelation
Paul
Paul

Reputation: 1117

visualization of correlation between replicates

To test the quality of the biological replicates by calculating Pearson correlation coefficient for the pairs of biological replicates for each cell-line.

A<-data.frame(A1=rnorm(100), A2=rnorm(100),
          A3=rnorm(100), B1=rnorm(100),
          B2=rnorm(100))

Some cases of the data have two replicates and others are three and contain no missing values. How to obtain such a plot to compare the replicates?

Quality of replicates

Upvotes: 0

Views: 1580

Answers (2)

mrip
mrip

Reputation: 15163

Here's one possible way. There's probably a more concise way to do this, though.

First thing, figure out which columns are replicates of which.

fullnames<-colnames(A)
basenames<-substr(fullnames,1,nchar(fullnames)-1)
repnum<-as.integer(substr(fullnames,nchar(fullnames),nchar(fullnames)))

Now compute the correlation matrix, and extract the data you need:

ca<-cor(A)
corMask<-upper.tri(ca) & basenames[col(ca)]==basenames[row(ca)]
corSub<-ca[corMask]
nameSub<-basenames[row(ca)[corMask]]
repnumSub<-apply(cbind(repnum[row(ca[corMask]],repnum[col(ca[corMask]]),1,paste,collapse="-")

Then draw the plot:

require(ggplot2)
plotdata<-data.frame(name=nameSub,cor=corSub,replicas=repnumSub)
ggplot(plotdata,aes(x=name,y=cor,pch=replicas))+geom_point()

Here's what it looks like, with the following sample data set:

set.seed(123)
A<-data.frame(A1=rnorm(100), A2=rnorm(100),A3=rnorm(100),
          B1=rnorm(100),B2=rnorm(100),
          C1=rnorm(100),C2=rnorm(100),C3=rnorm(100))

enter image description here

You can then add color or change the plot limits etc. to make it look the way you want.

Upvotes: 1

raymondowf
raymondowf

Reputation: 31

I suggest that the better representation would be using a heat map. The heat map visualizes the correlations between replicates. So, replicates from the same batch will show much higher correlation than other cell lines. So, ultimately you will see values close to 1.0 at the diagonal of the cells in your heat map. The heat map also shows the poor correlation between replicates from different cell lines. To perform such drawing function, you can use heatmap.2 from {gplot} package.

Upvotes: 0

Related Questions