library(MASS)


set.seed(123)

iter <- NULL
g1 <- NULL
tsum1 <- NULL
post1 <- NULL
tsum1i <- NULL
post1i <-NULL
g2 <- NULL
tsum2 <- NULL
post2 <- NULL
tsum2i <- NULL
post2i <- NULL

for(i in 1:500){
	# Randomly select 9 ALL cases and 3 AML cases

	r.chips <- c(sample(1:27, 9), sample(28:38, 3))  	

		train <- train[-r.chips,]
		indep <- train[r.chips,]

### One Gene Model Search ###
		grp <- c(rep(1, 18), rep(2, 8))
		gene <- NULL
		run1 <- SCVD(train)
		
	df.sort1 <- SortMat(Mat=run1, Sort=c(1,2))
						
		geneI <- as.numeric(row.names(df.sort1))[7129]
	g1[i] <- geneI	

	grp <- c(rep(1, 18), rep(2, 8))
	gene <- lXmat[,geneI]
	run2 <- SCVD(lXmat[,-geneI])
	
	df.sort2 <- SortMat(Mat=run2, Sort=c(1,2))
						
		bb <- as.numeric(row.names(df.sort2))[7128] 
			geneI.2 <- ifelse(bb < geneI, bb, bb+1)
	
iter[i] <- i

	grpf <- factor(grp, levels=1:2, labels = c("Positive", "Negative"))
		G1 <- matrix(lXmat[,geneI])

	f1 <- qda(G1, grpf, cv=T)
				
	tsum1[i] <- df.sort1[7129,1]
				
	post1[i] <-df.sort1[7129,2]
				
	# Check its performance on the independent set
		idata <- data.frame(indep[,geneI])
	ip1 <- predict(f1, idata)
			
	tsum1i[i] <- sum(ip1$class[1:9] == "Positive") + sum(ip1$class[10:12] == "Negative") 
				
	post1i[i] <- (prod(ip1$posterior[1:9])*prod(ip1$posterior[10:12]))	

		G2 <- train[,c(geneI,geneI.2)]
	f2 <- qda(G2, grpf, cv=T)
	p2 <- predict(f2)
				
	tsum2[i] <- df.sort2[7128,1]
				
	post2[i] <- df.sort2[7128,2]
				
	# Check its performance on the independent set
		idata2 <- data.frame(indep[,c(geneI,geneI.2)])
	ip2 <- predict(f2, idata2)
			
	tsum2i[i] <- sum(ip2$class[1:9] == "Positive") + sum(ip2$class[10:12] == "Negative") 
				
	post2i[i] <- (prod(ip2$posterior[1:9])*prod(ip2$posterior[10:12]))
}

out <- cbind(iter, g1, tsum1, post1, tsum1i, post1i, g2, tsum2, post2, tsum2i, post2i)

### To write this as a comma separated file use the following line.

write.table(out, file="out.csv", sep=",")