#This is version 2 of the similarity computations.  Whereas version 1 used the Masters et al. equation,
#version 2 uses a different equation defined by us in the manuscript

#Choose the input file
f = file.choose();

#Read the data
d=read.delim(f, colClasses=c(rep("character",3), rep("numeric", 30)))

cell.lines=d[,1]#The first column of d will be cell line names
dat=d[,2:33]#Columns 2 to 33 are the data for the 16 markers we will use.

#Convert X and Y in the first two columns of dat to be 0 and 1 respectively
temp=array(data=0, dim=dim(dat)[1])
q=which(dat[,1]=="Y")
temp[q]=1
dat[,1]=temp
temp=array(data=0, dim=dim(dat)[1])
q=which(dat[,2]=="Y")
temp[q]=1
dat[,2]=temp

#Find which row has MOLT-4 data
#q=grep(cell.lines, pattern="MOLT")
#Mark MOLT-4 data as missing
#dat[q,]=NA
#Convert dat from data.frame to matrix, so we can do operations across columns.
dat=as.matrix(dat)

num.lines=length(cell.lines)
#Create an empty array called "similarity" of size num.lines x num.lines to hold the similarity values
similarity=array(data=NA, dim=c(num.lines, num.lines))
#Give names to the rows and columns
colnames(similarity)=cell.lines
rownames(similarity)=cell.lines

#Compute similarities for each cell line pair
for(c1 in 1:num.lines)
{
	for(c2 in 1:num.lines)
	{
		cell.line1.data=dat[c1,]
		cell.line2.data=dat[c2,]
		T1=sum(cell.line1.data)
		T2=sum(cell.line2.data)
#Compute N, but only count instances where the STR difference is greater than 1
		N=sum(abs(cell.line1.data-cell.line2.data)>1)

		if(!is.na(T1) & !is.na(T2) & !is.na(N))#If all these values are good
		{
			similarity[c1,c2]=((T1+T2-N)/(T1+T2))*100
		}

	}
}

setwd(dirname(f));
#Write the similarity table to a CSV file
write.csv(similarity, file="similarity.csv", na="")