# Anne de Jong, May 2015, T-REx pipeline 
#
# 
# ----------------------------------- load the library--------------------------------------------------



library(statmod)
library(ggplot2)
library(reshape)
library(cluster)
library(plyr)
library(grid)
library(gplots)
library(igraph)
library(rjson)

library(limma)
library(edgeR)

lib_RNAseq <-  "T-REx_functions.R"

# ---------------------------------------  If you do not have the packages -------------------------------------------------------- 
# install.packages("ggplot2")
# install.packages("statmod")
# install.packages("reshape")
# install.packages("cluster")
# install.packages("plyr")
# install.packages("grid")
# install.packages("rjson")
# install.packages("igraph")
# install.packages("gplots")

# source("http://bioconductor.org/biocLite.R")
# biocLite()
# biocLite("limma")
# biocLite("edgeR")

# -------------------------------------------------------------------------------- for local use------------------------------------------------------ 
# sessiondir="E:\\Google Drive\\WERK\\Article - RNAseq\\Example_PNAS_CodY"
# sessiondir="E:\\tmp\\mirjam"
# sessiondir="E:\\tmp\\Example_PNAS_CodY"
# experiment="my_experiment"
# rpkm_file="RPKM.txt"
# factors_file="Factors.txt"
# class_file="Class.txt" 
# contrast_file="Contrasts.txt" 
# lib_RNAseq <-  "E:\\OwnCloud\\molgentools\\rnaseq\\g2d_RNAseq_multifactor_functions.R"


# -------------------------------------------------------------------------------- for webserver use ------------------------------------------------- 

# load parameters from command line or call from the webserver
# strange 'known!' thing with R is that is counts all the arguments, so inclusing the options :(
# R command line:  R --vanilla --slave --args . robyn -r bacillus_cereus_merged_normalized.rpkm -f Factors.txt -c Contrasts.txt < /usr/molgentools/rnaseq/RNAseq_multifactor.R
sessiondir <- commandArgs()[5]
experiment <- commandArgs()[6]
rpkm_file  <- commandArgs()[7]
factors_file  <- commandArgs()[8]
contrast_file <- commandArgs()[9]
class_file=commandArgs()[10]

# ----------------------------------- custom settings / or parameters from webserver--------------------------------------------------


# make web folder from the local folder
websessiondir <- gsub("/tmp/genome2d", "genome2d_results", sessiondir)

source(lib_RNAseq)
setwd(sessiondir)


# comment SINK if you run this script locally
sink("00.R.remarks.log", append=TRUE, split=FALSE)


# container for arguments and data tables in jSON format
jSON <- NULL
jSON$arguments['sessiondir'] 	<- sessiondir
jSON$arguments['experiment'] 	<- experiment
jSON$arguments['rpkm_file'] 	<- rpkm_file
jSON$arguments['factors_file'] 	<- factors_file
jSON$arguments['class_file']	<- class_file
jSON$arguments['contrast_file'] <- contrast_file

# graphics settings
pointsizes <- 48
# Image resolution

# 720p
x_720p <- 1280
y_720p <- 720
# FullHD
x_1k <- 1920
y_1k <- 1080
# QHD
x_2k <- 2560
y_2k <- 1440
# UHD 4k
x_4k <- 3840
y_4k <- 2160
# FUHD
x_8k <- 7680
y_8k <- 4320


 
 


logbook <- c()
logbook <- write_log(logbook, "Pipeline started")


#------------------------------------ Functions --------------------------------------------------

logbook <- write_log(logbook, "Functions loaded")

#------------------------------------ EdgeR --------------------------------------------------

## Multi Factorial design

# 1. read the factorial design
	logbook <- write_log(logbook, "# 1. Read the factorial design")
	MF <- read.delim(factors_file,sep="\t")
	MF

	# generic import of undefined number of factors
	MF.dim <- dim(MF)[2] 
	MF.factors <- MF[,2] # first factor
	# add more factors if there are any (MF.dim >2)
	if (MF.dim >2) {
		for (i in 3:MF.dim) { MF.factors <- paste(MF.factors,MF[,i],sep=".") }
	}
 	MF.group <- factor(MF.factors)
	MF.group.all <- cbind(MF,Group=MF.group)
	MF.group.levels <- as.list(levels(MF.group.all$Group))
	MF.group.all


# 2. Read the RNAseq data
	logbook <- write_log(logbook, "# 2. Read the RNAseq data")
	x <- read.delim(rpkm_file,sep="\t", row.names="key")
	y <- DGEList(counts=x,group=MF.group)
	head(x)
	
	x.melt <- melt(x)
	colnames(x.melt) =c("Experiment","Value")
	x.melt$log2Value <- log(x.melt$Value,2)

	my_barplot_palette <- colorRampPalette(c("#B4AF91",  "#C03000"))(n = length(x))
	filename <- paste(experiment, "colored_boxplot","png", sep=".")
	png(file = filename, width = x_1k, height = y_1k, units = "px", pointsize = 12, bg = "white")	
	p <- ggplot(x.melt,  aes(Experiment,log2Value) ) +	geom_boxplot( aes(fill=Experiment)) + ggtitle("Boxplot of Expression Values of Experiments") + theme_g2d_barplot +	theme(legend.position="none") +	scale_fill_manual(values = my_barplot_palette)
	print(p)
	dev.off()
	
# 3. Read the class file
	logbook <- write_log(logbook, "# 3. Read the class file")
	my_class_genes <- read.delim(class_file,sep="\t")
	colnames(my_class_genes) =c("GeneID","color","Group")
	

# 4. Create the design matrix
	logbook <- write_log(logbook, "# 4. Create the design matrix")
	designMF <- model.matrix(~0+MF.group,  data=y$samples)
	designMF
	colnames(designMF) <- levels(MF.group)
	designMF

# 5. Normalization 
	logbook <- write_log(logbook, "# 5. Normalization ")
	# raw data
	filename <- paste(experiment, "Box_plot_raw","png", sep=".")
	png(file = filename, width = x_4k, height = y_4k, units = "px", pointsize = 12, bg = "white")
	boxplot(as.data.frame(log(y$counts,2)), main="RPKM raw data (log2)", ylab="log2(counts)",  notch=FALSE)
	dev.off()

	y<-calcNormFactors(y)
	
# 6. Calculate mean values of the replicates
	logbook <- write_log(logbook, "# 6. Calculate mean values of the replicates")

	# Take mean of replicates if are are replicates 
	y_groupheader <- y$counts
	colnames(y_groupheader)<-MF.group 
	mean_table <- c()
	for (i in MF.group.levels) {
		print(i)
		my_cols<-which(colnames(y_groupheader) %in% i)
		print(my_cols)
		if (length(my_cols)>1) {
			my_mean<-rowMeans(y_groupheader[,my_cols])
		} else {
			my_mean<-(y_groupheader[,my_cols])
		}	
		mean_table<-cbind(mean_table,my_mean)
	}

	colnames(mean_table)<-MF.group.levels
	head(mean_table)
	filename <- paste(experiment, "Mean_signals_of_replicates.txt", sep=".")
	write.table(mean_table, file=filename , quote=F, sep="\t", col.names=NA)
	jSON$Mean_signals_of_replicates <- filename

	
	# Mean Signals Normalized per gene for heatmaps  (x-min)/max
	mean_table_log <- log(mean_table,2)
	head(mean_table_log)
	filename <- paste(experiment, "Mean_signals_log2.txt", sep=".")
	write.table(mean_table_log, file=filename , quote=F, sep="\t", col.names=NA)	
	jSON$Mean_signals_log2 <- filename
	
	mean_table_log_scaled <- round((100 * mean_table_log / max(mean_table_log)), digits=3)
	head(mean_table_log_scaled)
	filename <- paste(experiment, "Mean_signals_log2_scaled.txt", sep=".")
	write.table(mean_table_log_scaled, file=filename , quote=F, sep="\t", col.names=NA)
	jSON$Mean_signals_log2_scaled <- filename
	
	mean_table_log_scaled_per_gene <- round(t(apply(mean_table_log, 1, function(x) (x-min(x))/(max(x)-min(x)))), digits=4)
	filename <- paste(experiment, "Mean_signals_log2_scaled_per_gene.txt", sep=".")
	write.table(mean_table_log_scaled_per_gene, file=filename , quote=F, sep="\t", col.names=NA)
	jSON$Mean_signals_log2_scaled_per_gene <- filename
	
	# add class (color and groups) to the mean table
	mean_table_log_scaled_class <- mean_table_log_scaled
	mean_table_log_scaled_class <- cbind(mean_table_log_scaled_class, row.names(mean_table_log_scaled_class))
	colnames(mean_table_log_scaled_class) <- c(colnames(mean_table_log_scaled), "GeneID")
	mean_table_log_scaled_class <- merge(mean_table_log_scaled_class, my_class_genes, by="GeneID",all.x=T)
	mean_table_log_scaled_class$color <- ifelse(is.na(mean_table_log_scaled_class$color), "grey", as.character(mean_table_log_scaled_class$color))	# add a default color if the color is NA
	mean_table_log_scaled_class$Group <- ifelse(is.na(mean_table_log_scaled_class$Group), "NA", as.character(mean_table_log_scaled_class$Group))		# add a default Group if the Group is NA
	head(mean_table_log_scaled_class)
	filename <- paste(experiment, "Mean_signals_log2_scaled_CLASSES.txt", sep=".")
	write.table(mean_table_log_scaled_class, file=filename , quote=F, sep="\t", col.names=NA)
	jSON$Mean_signals_log2_scaled_CLASSES <- filename
	

# 7. Graphics for each subClass on the basis of the group column
	logbook <- write_log(logbook, "# 7. Signal plots of Class genes")
	mean_table_class <- subset(mean_table, row.names(mean_table) %in% my_class_genes$GeneID)
	mean_table_class_log <- subset(mean_table_log, row.names(mean_table_log) %in% my_class_genes$GeneID)
	head(mean_table_class)
	if (nrow(mean_table_class)<5) {
		logbook <- write_log(logbook, "# ERROR - 7. Skipping Class signal plots due to low numbers classes found")
	} else {
		#draw a graph for each class color
		logbook <- write_log(logbook, "-   a) Create Graphics for mean signals of each subClass")
		Class_signal_plots(mean_table_class, "Mean_Signals_subClass")
		logbook <- write_log(logbook, "-   b) Create Graphics for mean log2(signal) of each subClass")
		Class_signal_plots(mean_table_class_log, "Mean_Signals_subClass_log2")
	}

# 8. Correlation matrix of All Class genes, using mean signal of replicates
	logbook <- write_log(logbook, "# 8. Correlation matrix of All Class genes, using mean signal of replicates")
	cor_mean_table_class <- cor(t(mean_table_class))	
	filename <- paste(experiment, "Correlation_matrix_Class_genes.txt", sep=".")
	write.table(cor_mean_table_class, file=filename , quote=F, sep="\t", row.names=F)
	jSON$Correlation_matrix_Class_genes <- filename

	filename <- paste(experiment, "Correlation_matrix_Class_genes.png", sep=".")
	png(file = filename, width = x_2k, height = y_2k, units = "px", pointsize = 12, bg = "white")
	m1 <- melt(cor_mean_table_class, id = row.names)
	qplot(x=X1, y=X2, data=as.data.frame(m1),main = "Correlation matrix of Class genes", xlab="Class genes", ylab="Class genes", fill=value, geom="tile", zlim= c(0,1)) + theme_g2d_plot_cor + scale_fill_gradient2(limits=c(-1, 1))
	dev.off()

# 9. Correlation matrix of subClass genes, using mean signal of replicates
	logbook <- write_log(logbook, "# 9. Correlation matrix of subClass genes, using mean signal of replicates")
	SubClass_correlation_plots(cor_mean_table_class, "SubClass_correlation_matrix" )

	
# 10. MDS plot
	logbook <- write_log(logbook, "# 10. MDS plot")
	filename <- paste(experiment, "MDS.plot","png", sep=".")
	png(file = filename, width = x_1k, height = y_1k, units = "px", pointsize = 12, bg = "white")
	par(mar=c(6,8,3,6)) # bottom, left, top, right
	plotMDS(y, col="blue", top=500, main="MDS Plot of Experiments", cex.main=2,cex.axis=1.8, cex.lab=2.5, cex=3 )
	dev.off()	

# 11. Library size
	logbook <- write_log(logbook, "# 11. Library size")
		# old version par(mar=c(6,6,3,16)) # bottom, left, top, right
		# old version my_barplot_palette <- colorRampPalette(c("#B4AF91", "#787746", "#40411E", "#32331D", "#C03000"))(n = 10)
		# old version barplot(y$samples$lib.size*1e-6, legend =y$samples$group, names=row.names(y$samples), xlab="Samples", ylab="Depth (millions)", col=my_barplot_palette, cex.main=2,cex.axis=1.8,cex.lab=2.5, main="Library Size", beside = TRUE, )

	# prepare data for barplot
	my_samples <- y$samples
	my_samples$experiment <- row.names(my_samples)
	my_samples$lib.size <- as.numeric(my_samples$lib.size)
	my_samples$experiment <-factor(my_samples$experiment, levels=my_samples[order(my_samples$experiment), "experiment"])
	#my_barplot_palette <- colorRampPalette(c("#0057FF", "#001640"))(n = dim(my_samples)[1])
	my_barplot_palette <- colorRampPalette(c("#B4AF91", "#787746", "#40411E", "#32331D", "#C03000"))(n = dim(my_samples)[1])
	bar_chart <- ggplot(my_samples, aes(experiment, lib.size, fill=factor(lib.size))) +
			ggtitle("Library Sizes") + 
			theme_g2d_barplot + 
			geom_bar(stat="identity") +
			xlab("Experiments") +  
			ylab("Number of reads") +
			scale_fill_manual(values = my_barplot_palette) +
			theme_g2d_barplot +
			theme(legend.position="none") +
			scale_y_continuous(expand = c(0,0))
	filename <- paste(experiment, "Library_size","png", sep=".")
	png(file = filename, width = x_720p, height = y_720p, units = "px", pointsize = 12, bg = "white")
	print(bar_chart)
	dev.off()	
	
# 12. Calculation of Dispersion
	logbook <- write_log(logbook, "# 12. Calculation of Dispersion")
	y <- estimateGLMCommonDisp(y,designMF)
	y <- estimateGLMTrendedDisp(y,designMF, method="bin.loess")
	y <- estimateGLMTagwiseDisp(y,designMF)

# 13. Load the Contrasts
	logbook <- write_log(logbook, "# 13. Load the Contrasts")
	# contrast from file
	my_contrasts <- readLines(contrast_file)
	my_contrasts
	contrast.matrix <- makeContrasts(contrasts=my_contrasts, levels=designMF)
	contrast.matrix
	contrast.matrix.dim <- dim(contrast.matrix)[2]

# 14. Fit the data
	logbook <- write_log(logbook, "# 14. Fit the data")
	fit.design <- glmFit(y, designMF)
	lrt.design <- glmLRT(fit.design, designMF)
	filename <- paste(experiment, "Normalised_data_table.txt", sep=".")
	write.table(lrt.design$fitted.values, file=filename , quote=F, sep="\t", col.names=NA)
	jSON$Normalised_data_table <- filename

	
# 15. Make correlation Matrix of experiments
	logbook <- write_log(logbook, "# 15. Make correlation Matrix of experiment")
	logFC <- predFC(y,designMF,prior.count=1,dispersion=0.05)
	corLogFC <- cor(logFC)
	filename <- paste(experiment, "Correlation_matrix_experiments.txt", sep=".")
	write.table(corLogFC, file=filename , quote=F, sep="\t", row.names=F)	
	jSON$Correlation_matrix_experiments <- filename

	filename <- paste(experiment, "Correlation_matrix_experiments.png", sep=".")
	png(file = filename, width = x_1k, height = y_1k, units = "px", pointsize = 12, bg = "white")
 	m1 <- melt(corLogFC, id = row.names)
	qplot(x=X1, y=X2, data=as.data.frame(m1),main = "Correlation matrix experiments", xlab="Experiments", ylab="Experiments", fill=value, geom="tile") + theme_g2d_plot_cor
	dev.off()	

# 16. Loop through the Contrasts and draw for each Contrast a: 
		## - DE (Differential Expression ) table
		## - Significant Fold Change Plots
		## - MA plot
		## - Volcano plot
		## - Fold change distribution plot
		## - Count distribution plot
	logbook <- write_log(logbook, "# 16. Loop through the Contrasts")
	tophits <- c()
	tophits_highfold <-c()
	tophits_no_background <-c()
	filenames.MAplots <- list()
	filenames.VolcanoPlots <- list()
	filenames.SigChanged_plots <- list()
	filenames.Significant_Changed_Genes_tables <- list()
	my_overview <- c()
	my_GeneNetwork <- c()
	volcano_tables <- NULL

	for (i in 1:contrast.matrix.dim) {
		logbook <- write_log(logbook, my_contrasts[i] )
		lrt <- glmLRT(fit.design , contrast=contrast.matrix[,i])
		my.DGEList.glmFit = topTags(lrt, n=100000, adjust.method="BH")
	# A. DE (Differential Expression ) table
		output = cbind(rownames(my.DGEList.glmFit$table), my.DGEList.glmFit$table)
		# add fold change
		output <- cbind(output, sign(output$logFC)*(2^abs(output$logFC)))
		colnames(output) =c("GeneID","logFC", "logCPM", "LR", "pvalue", "adj_pvalue", "Fold")
		output$minFDR <- -log(output$adj_pvalue,2)
		filename <- paste(experiment, "Differential_Expression", i, my_contrasts[i] ,"txt", sep=".")
		write.table(output, file=filename , quote=F, sep="\t", row.names=F)
		jSON$Differential_Expression[my_contrasts[i]] <- filename

		
	# B. For the html table only export the up and down regulated genes
		tophits_ONLY <- output[( abs(output$Fold)>=2 & output$adj_pvalue<0.05),]
		tophits_ONLY <- tophits_ONLY[with(tophits_ONLY, order(-Fold)), ]
		tophits_ONLY$logFC <- sprintf("%.2f", tophits_ONLY$logFC)
		tophits_ONLY$logCPM <- sprintf("%.2f", tophits_ONLY$logCPM)
		tophits_ONLY$LR <- sprintf("%.1f", tophits_ONLY$LR)
		#tophits_ONLY$minFDR <- -log(tophits_ONLY$adj_pvalue,2)
		tophits_ONLY$minFDR <- sprintf("%.2f", tophits_ONLY$minFDR)
		tophits_ONLY$pvalue <- sprintf("%.1e", tophits_ONLY$pvalue) 
		tophits_ONLY$adj_pvalue <- sprintf("%.1e", tophits_ONLY$adj_pvalue) 
		tophits_ONLY$Fold <- sprintf("%.1f", tophits_ONLY$Fold)
		filename <- paste(experiment, "Significant_Changed_Genes", i, my_contrasts[i] ,"txt", sep=".")
		filenames.Significant_Changed_Genes_tables[i] <- filename	
		write.table(tophits_ONLY, file=filename , quote=F, sep="\t", row.names=F)	

	# C. GeneNetwork; Add tophits for each contrast to network compatible table
		my_GeneNetwork <- rbind(my_GeneNetwork, data.frame(Contrasts=my_contrasts[i], GeneID=tophits_ONLY$GeneID, logFC=tophits_ONLY$logFC, adj_pvalue=tophits_ONLY$adj_pvalue))
		

	# E. Add data to overview table
		total<- length(output$GeneID)
		up   <- length(output[( output$Fold >= 2 & output$adj_pvalue <= 0.05),]$GeneID)
		down <- length(output[( output$Fold <= -2 & output$adj_pvalue <= 0.05),]$GeneID)
		my_overview <- rbind(my_overview, data.frame(Contrasts=my_contrasts[i], Total=total, Up=up, Down=down, TableFilename=filename))	

		
	# D. Plot tophits_ONLY for all genes
		filename <- paste(experiment, "Significant_Changed_Genes", i, my_contrasts[i] ,"png", sep=".")
		filenames.SigChanged_plots[i] <- filename		
		png(file = filename, width = x_2k, height = y_2k, units = "px", pointsize = 12, bg = "white")
		my_title=paste(experiment, "Significant changed genes", my_contrasts[i], sep=" ")
		tophits_ONLY$logFC <- as.numeric(tophits_ONLY$logFC)
		tophits_ONLY$GeneID <-factor(tophits_ONLY$GeneID, levels=tophits_ONLY[order(-tophits_ONLY$logFC), "GeneID"])
		tophits_ONLY$Up_Down <- ifelse(tophits_ONLY$logFC>0, "UP", "DOWN")
		bar_chart <- ggplot(tophits_ONLY, aes(GeneID, logFC, fill=factor(Up_Down), colour = factor(Up_Down)), main = my_title ) + 
			ggtitle(my_title) + 
			theme_g2d_barplot + 
			geom_bar(stat="identity") +  
			xlab("Genes") +  ylab("Fold Change (Log2FC)") + 
			scale_colour_manual(breaks = tophits_ONLY$updown, values = c("#FFFAF0","#5494FF")) + 
			scale_fill_manual(breaks = tophits_ONLY$updown, values = c("#FFD312","#0608B2"))
		print(bar_chart)
		dev.off()
	
	# E. Plot All genes of each class
		# to do
		
		
	# F. add the GeneIDs to the Tophits list for all experiments
		tophits <- c(tophits, as.character(output[( abs(output$Fold)>2 & output$adj_pvalue<0.05),]$GeneID))
		tophits_highfold <- c(tophits_highfold, as.character(output[( abs(output$Fold)>5 & output$adj_pvalue<0.01),]$GeneID))
		tophits_no_background <- c(tophits_no_background, as.character(output[( abs(output$Fold)>1.4 & output$adj_pvalue<0.20),]$GeneID))

		
		
	# G. Add all individual files to one file for TMEV or other downstream programs	
		my_columns <- cbind(rownames(output), output$logFC)
		colnames(my_columns) = c("GeneID","logFC")
		if (i == 1 ) {
			TMEV <- my_columns
			TMEV.colnames <- c("GeneID",my_contrasts[i]) ;
		}
		if (i > 1 ) {	
			TMEV<-merge(TMEV, my_columns, by="GeneID",all.x=T)
			TMEV.colnames <- c(TMEV.colnames,my_contrasts[i]) ;
		}	
		colnames(TMEV) = TMEV.colnames

	# H. Select data for colouring classes 
		my_class <- output[which(row.names(output) %in% my_class_genes$GeneID),]
		my_class <- merge(my_class, my_class_genes,  by="GeneID",all.x=T)
		colnames(my_class) =c("GeneID","logFC", "logCPM", "LR", "pvalue", "adj_pvalue", "Fold", "minFDR", "color")
		# Select data with 0 expression and mark with X it in the plots
   		my_null_class <- output[which(row.names(output) %in% row.names(lrt$fitted.values)[lrt$fitted.values==0]),]
		my_null_class
		
	# I. - MA plots
		filename <- paste(experiment, "MAplot", i, my_contrasts[i] ,"png", sep=".")
		filenames.MAplots[i] <- filename
		plottitle <- paste("MAplot", experiment, "contrast=", my_contrasts[i] , sep=" ")
		png(file = filename, width = x_1k, height = y_1k,  units = "px", pointsize = 12, bg = "white")	
		par(mar=c(6,6,3,2)) # bottom, left, top, right
		plot(my.DGEList.glmFit$table$logCPM, my.DGEList.glmFit$table$logFC, col="grey39", cex.main=2,cex.axis=1.8,cex.lab=2.5, main=plottitle, pch=20, ylab="M - log2 ratio", xlab="A - mean expression level")
		points(my_class$logCPM, my_class$logFC, pch=19,lwd=1.2, col=as.character(my_class$color))
		points(my_null_class$logCPM, my_null_class$logFC, pch=4,lwd=1, col="blue")

		rect(-100,-1, 100,1, col = "blue", density = 10, border = "transparent")
		text(output$logCPM[1:20], output$logFC[1:20], labels = row.names(output[1:20,]), pos = 4, cex=0.8, col="blue")
		abline(h=0)
		dev.off()
		
	# J - Volcano Data
		# Export Volcano data including class data and mean signal
		volcano_table <- cbind(row.names(output), output$logFC, output$minFDR, output$logCPM)
		colnames(volcano_table) <- c("GeneID", "logFC", "minFDR", "logCPM")
		volcano_table <- merge(volcano_table, my_class_genes, by="GeneID",all.x=T)
		volcano_table$color <- ifelse(is.na(volcano_table$color), "grey", as.character(volcano_table$color))	# add a default color if the color is NA
		volcano_table$Group <- ifelse(is.na(volcano_table$Group), "NA", as.character(volcano_table$Group))		# add a default Group if the Group is NA
		volcano_table$minFDR <- as.numeric(as.character(volcano_table$minFDR))
		volcano_table <- volcano_table[ order(volcano_table$minFDR,decreasing = TRUE), ]
		# add signal data and scale it to color all above the median
		volcano_table$logCPM <- as.numeric(as.character(volcano_table$logCPM))
		volcano_table$sfere <- volcano_table$logCPM - median(volcano_table$logCPM)
		volcano_table$sfere <- ifelse(volcano_table$sfere<0, 0, volcano_table$sfere)
		head(volcano_table)
		filename  <- paste(experiment, "Volcano_plot", i, my_contrasts[i] ,"txt", sep=".")
		write.table(volcano_table, file=filename , quote=F, sep="\t", col.names=NA)

		
		
		jSON$VolcanoPlots[my_contrasts[i]] <- filename
		# outside the loop we want to filter on tophits_no_background, so we store this table
		volcano_tables[i] <- list(volcano_table)
		
		# Volcano Plot
		filename  <- paste(experiment, "Volcano_plot", i, my_contrasts[i] ,"png", sep=".")
		filenames.VolcanoPlots[i] <- filename
		plottitle <- paste("Volcano plot of", experiment, "contrast=", my_contrasts[i] , sep=" ")
		png(file = filename, width = x_1k, height = y_1k, units = "px", pointsize = 12, bg = "white")	
		par(mar=c(6,6,3,2)) # bottom, left, top, right
		plot(my.DGEList.glmFit$table$logFC, -log(my.DGEList.glmFit$table$FDR,2), col="grey39", cex.main=2,cex.axis=1.8,cex.lab=2.5, main=plottitle, pch=20, xlab="log2 ratio", ylab="-log2(FDR p-value)")
		points(my_class$logFC, -log(my_class$adj_pvalue,2), pch=19,lwd=1.2, col=as.character(my_class$color))
		points(my_null_class$logFC, -log(my_null_class$adj_pvalue,2), pch=4,lwd=1, col="blue")
		text(output$logFC[1:10], -log(output$adj_pvalue[1:10],2), labels = row.names(output[1:10,]), pos = 3, cex=0.9, col="blue")
		lines(c(-1,-1),c(-100,1000))
		lines(c(1,1),c(-100,1000))
		lines(c(-3.3,-3.3),c(-100,1000), lty=2)
		lines(c(3.3,3.3),c(-100,1000), lty=2)
		lines(c(-100,100),c(4.4,4.4))
		lines(c(-100,100),c(6.6,6.6), lty=2)
		rect(-100,-5,100,4.1, col = "red", density = 10, border = "transparent")
		rect(-0.95,-1,0.95,1000, col = "red", density = 10, border = "transparent")
		dev.off()
		
	# K. - Fold change distribution plots
		filename  <- paste(experiment, "Fold_distribution", i, my_contrasts[i] ,"png", sep=".")
		plottitle <- paste("Fold change distribution", experiment, "contrast=", my_contrasts[i] , sep=" ")
		png(file = filename, width = 1280, height = 1024, units = "px", pointsize = 12, bg = "white")	
		par(mar=c(6,6,3,2)) # bottom, left, top, right
		plot(rev(output$logFC), cex.main=2,cex.axis=1.8,cex.lab=2.5, main=plottitle, pch=20, xlab="genes (sorted on p-value)", ylab="log2(Fold Change)")
		rect(0,-1, nrow(output),1, col = "red", density = 30, border = "transparent")
		dev.off()

	# L - Count distribution plots
		filename  <- paste(experiment, "Count_distribution", i, my_contrasts[i] ,"png", sep=".")
		plottitle <- paste("Count distribution", experiment, "contrast=", my_contrasts[i] , sep=" ")
		png(file = filename, width = 1280, height = 1024, units = "px", pointsize = 12, bg = "white")	
		par(mar=c(6,6,3,2)) # bottom, left, top, right
		plot(sort(output$logCPM), cex.main=2,cex.axis=1.8,cex.lab=2.5, main=plottitle, pch=20, xlab="genes used for normalization", ylab="log2(Counts)")
		rect(0,-100, 0.2*nrow(output),100, col = "red", density = 30, border = "transparent")
		rect(0.8*nrow(output),-100, nrow(output),100, col = "red", density = 30, border = "transparent")
		dev.off()		
		
	}
	# save filenames
	filename <- paste(experiment, "MAplots.filenames.txt", sep=".")
	write.csv(filenames.MAplots, file=filename)
	filename <- paste(experiment, "VolcanoPlots.filenames.txt", sep=".")
	write.csv(filenames.VolcanoPlots, file=filename)
	filename <- paste(experiment, "SigChanged_plots.filenames.txt", sep=".")
	write.csv(filenames.SigChanged_plots, file=filename)
	filename <- paste(experiment, "Significant_Changed_Genes_tables.filenames.txt", sep=".")
	write.csv(filenames.Significant_Changed_Genes_tables, file=filename)
	
	# save my_overview table
	filename <- paste(experiment, "Contrasts_overview.txt", sep=".")
	write.table(my_overview, file=filename , quote=F, sep="\t", row.names=FALSE)
	jSON$Contrasts_overview <- filename	

	# Store data for volcano plots where at least one gene is above the background
	logbook <- write_log(logbook, "# 16.5 Write filtered Volcano plots data" )
	
	# Create a items list containing the plots 
	items <- list()
	head(as.data.frame(volcano_tables))
	for (i in 1:contrast.matrix.dim) {
		volcano_filtered <- as.data.frame(volcano_tables[i])
		volcano_filtered <- volcano_filtered[which(volcano_filtered$GeneID %in% tophits_no_background),]
		filename  <- paste(experiment, "Volcano_plot_filtered", i, my_contrasts[i] ,"txt", sep=".")
		write.table(volcano_filtered, file=filename , quote=F, sep="\t", col.names=NA)
		#jSON$VolcanoPlotsFiltered[my_contrasts[i]] <- filename
		
		# JSON: Description of each plot
		item <- NULL
		item$name <- my_contrasts[i]  
		item$type <- "scatterplot"
		item$settingsFile <- 'item-settings/molgen-volcanoplot.json'
			settings <- NULL
			settings$filename <-  paste(websessiondir, filename, sep="/")
		item$settings <- toJSON(settings)	
		# add the description to the items list
		items <- c(items, toJSON(item))
	}	
	
# 17. Draw the network using iGraph
	logbook <- write_log(logbook, "# 17. Draw Networks" )
	
	# The complete network
	filename <- paste(experiment, "GeneNetwork_EdgeList.txt", sep=".")
	write.table(my_GeneNetwork, file=filename , quote=F, sep="\t", row.names=FALSE)	
	jSON$GeneNetwork_EdgeList <- filename
	
	my_Error <- tryCatch(
      Draw_GeneNetwork(my_GeneNetwork, "GeneNetwork_of_Contrasts", layout.reingold.tilford ),
      error=function(e) e
	)
	

	# The high fold change network
	my_Error <- tryCatch(
	{
		my_GeneNetwork_high <- my_GeneNetwork[( abs(as.numeric(as.character(my_GeneNetwork$logFC)))>=4 & as.numeric(as.character(my_GeneNetwork$adj_pvalue))<0.01 ),]
		# frequency 
		freq_table <- table(as.character(my_GeneNetwork_high$GeneID))
		
		filename <- paste(experiment, "GeneNetwork_EdgeList_high.txt", sep=".")
		write.table(my_GeneNetwork_high, file=filename , quote=F, sep="\t", row.names=FALSE)
		jSON$GeneNetwork_EdgeList_high <- filename
		Draw_GeneNetwork(my_GeneNetwork_high, "GeneNetwork_of_Contrasts_high", layout.reingold.tilford )
	},
		error=function(e) e
	)


		
	
# 18. TMEV: Combine the output of all the Contrasts to one file that is compatible with TMEV (TIGR Multi Experiment Viewer)
	logbook <- write_log(logbook, "# 18. Combine the output of all the Contrasts to one file that is compatible with TMEV" )
	colnames(TMEV) = TMEV.colnames
	
	filename <- paste(experiment, "TMEV", "txt", sep=".")
	write.table(TMEV, file=filename , quote=F, sep="\t", row.names=F)
	
	row.names(TMEV) <- TMEV$GeneID
	TMEV[1] <- NULL
	TMEV_TOPHITS <- TMEV[which(row.names(TMEV) %in% tophits),]
	TMEV_TOPHITS <- TMEV_TOPHITS[ order(TMEV_TOPHITS[,1],decreasing = TRUE), ]
	TMEV_TOPHITS_highfold <- TMEV[which(row.names(TMEV) %in% tophits_highfold),]
	TMEV_TOPHITS_highfold <- TMEV_TOPHITS_highfold[ order(TMEV_TOPHITS_highfold[,1],decreasing = TRUE), ]

	# To do: Expression profile plots of class genes and tophit genes
	#CLASS_expression <- y[which(row.names(y) %in% tophits),]
	#CLASS_expression <- y[ order(TOPHITS_expression[,1],decreasing = TRUE), ]
	#TOPHITS_expression <- y[which(row.names(y) %in% tophits),]
	#TOPHITS_expression <- y[ order(TOPHITS_expression[,1],decreasing = TRUE), ]
	#TOPHITS_expression_highfold <- y[which(row.names(y) %in% tophits_highfold),]
	#TOPHITS_expression_highfold <- y[ order(TOPHITS_expression_highfold[,1],decreasing = TRUE), ]

		
	filename <- paste(experiment, "TMEV_TOPHITS_logFC", "txt", sep=".")
	write.table(TMEV_TOPHITS, file=filename , quote=F, sep="\t", row.names=T, col.names=NA)	
	filename <- paste(experiment, "TMEV_TOPHITS_logFC_highfold", "txt", sep=".")
	write.table(TMEV_TOPHITS_highfold, file=filename , quote=F, sep="\t", row.names=T, col.names=NA)

# ------------------------------------------------------------------------------------------------------------ heatmaps -------------------------------------------------------------------------------------------

# Heatmap settings
	my_heatmap_palette_signal <- colorRampPalette(c("#FCFFF5", "#D1DBBD", "#91AA9D", "#3E606F", "#193441"))(n = 1000) 
	my_heatmap_palette_ratio <- colorRampPalette(c("#4C3200", "#CC8400", "white","#00006B", "#00004C"))(n = 1000)  # orange - blue
	my_heatmap_margins <- c(14,10)
	my_heatmap_pointsize <- 24


# 19. Heatmap of experiments to see how replicates perform
	logbook <- write_log(logbook, "# 19. Heatmaps" )
	
	n <- length(row.names(lrt.design$fitted.values))
	normalized_signals <- as.data.frame((lrt.design$fitted.values))
	normalized_signals <- normalized_signals[order(normalized_signals[,1]),]
	normalized_signals <- normalized_signals[round(n*.1):round(n-n*.3),]	# remove high and low 20% counts
	# save the heatmap data to file
		filename <- paste(experiment, "Heatmap_EXPERIMENTS", "txt", sep=".")
		write.table(normalized_signals, file=filename , quote=F, sep="\t", row.names=T, col.names=NA)	
	filename <- paste(experiment, "Heatmap_EXPERIMENTS", "png", sep=".")
	png(file = filename, width = x_2k, height = y_2k, units = "px", pointsize = my_heatmap_pointsize, bg = "white")	
	par(mar=c(15,6,13,2)) # bottom, left, top, right
	heatmap.2(as.matrix(log2(normalized_signals+0.5)),dendrogram="both",Rowv=TRUE, col=my_heatmap_palette_signal, scale="none", key=T, keysize=1, main="Normalized log2(Signals), Experiments vs Genes",margins=my_heatmap_margins, density.info="none", trace="none",cexCol=0.9, labRow=NA)
	dev.off()
	

	
	
# 20. Heatmaps of TopHits
	logbook <- write_log(logbook, "# 20. Draw heatmaps of TopHits" )

	heatmap_temp <-(t(apply(TMEV_TOPHITS,1,as.numeric)))
	colnames(heatmap_temp) <- colnames(TMEV_TOPHITS)
	Anne_heatmap(heatmap_temp, experiment, "Heatmap of TopHits Ratio (logFC)", "Heatmap_TopHits", "ratio")

	# Heatmap High fold change
	heatmap_temp <-(t(apply(TMEV_TOPHITS_highfold,1,as.numeric)))
	colnames(heatmap_temp) <- colnames(TMEV_TOPHITS_highfold)	
	Anne_heatmap(heatmap_temp, experiment, "Heatmap of High Fold Changed TopHits", "Heatmap_TopHits_HighFold", "ratio")

	
	

# 21. Heatmap: Combine Class data with TopHits data and draw plots of sub classes
	logbook <- write_log(logbook, "# 21.  Combine Class data with TopHits data and draw plots and heatmaps of sub classes" )
	CLASS_TOPHITS <- subset(TMEV_TOPHITS, row.names(TMEV_TOPHITS) %in% my_class_genes$GeneID)
	heatmap_temp <-(t(apply(CLASS_TOPHITS,1,as.numeric)))
	colnames(heatmap_temp) <- colnames(CLASS_TOPHITS)	
	Anne_heatmap(heatmap_temp, experiment, "Heatmap of TopHits of all Class genes (logFC)", "Heatmap_Class_TopHits", "ratio")

# 22. Heatmap of each Class Ratio
	logbook <- write_log(logbook, "# 22. Make Ratio Heatmap of each Class" )
	Class_GroupNames <- levels(my_class_genes$Group)
	filenames.ClassHeatmaps <- list()
	filenames.Clusters <- list()
	for (i in 1:length(Class_GroupNames)) {
		Class_Genes <- subset(my_class_genes$GeneID, my_class_genes$Group %in% Class_GroupNames[i])
		Class_Group <- subset(TMEV, row.names(TMEV) %in% Class_Genes)
		if (dim(Class_Group)[1] > 1) {
			filename_prefix   <- paste("Heatmap_Class", Class_GroupNames[i], sep=".")
			filenames.ClassHeatmaps <- c(filenames.ClassHeatmaps, paste(experiment, filename_prefix ,"png", sep="."))
			filenames.Clusters  <- c(filenames.Clusters, paste(experiment, filename_prefix ,"clusters", "txt", sep="."))
			main_title <- paste(experiment, "Heatmap of Contrasts of Class", Class_GroupNames[i] , sep=" ")
			heatmap_temp <-(t(apply(Class_Group,1,as.numeric)))
			colnames(heatmap_temp) <- colnames(Class_Group)	
			Anne_heatmap(heatmap_temp, experiment, main_title, filename_prefix, "ratio")
		}	
	}
	# save filenames
	filename <- paste(experiment, "ClassHeatmaps.filenames.txt", sep=".")
	write.csv(filenames.ClassHeatmaps, file=filename)
	filename <- paste(experiment, "ClassHeatmaps.clusters.filenames.txt", sep=".")
	#write.table(unlist(filenames.ClassHeatmaps), file=filename, row.names = FALSE, col.names = FALSE, quote=FALSE)
	temp <- NULL
	temp$files <- sapply(filenames.Clusters, paste0, collapse=",") 
	temp$names <- sapply(Class_GroupNames, paste0, collapse=",") 
	write.table(as.data.frame(temp), file=filename, row.names = FALSE, col.names = FALSE, quote=FALSE)
	

	
# 23. High Fold change
	logbook <- write_log(logbook, "# 23. High Fold changed genes" )
	CLASS_TOPHITS_highfold <- subset(TMEV_TOPHITS_highfold, row.names(TMEV_TOPHITS) %in% my_class_genes$GeneID)
	filename <- paste(experiment, "CLASS_TOPHITS_highfold_logFC", "txt", sep=".")
	write.table(CLASS_TOPHITS_highfold, file=filename , quote=F, sep="\t", row.names=T, col.names=NA)	

# 24. Heatmaps of Class Signal data
	logbook <- write_log(logbook, "# 24. Signal data Heat Maps" )
	CLASS_TOPHITS_signal <- subset(mean_table_class, row.names(mean_table_class) %in% tophits)
	heatmap_temp <-(t(apply(CLASS_TOPHITS_signal,1,as.numeric)))
	colnames(heatmap_temp) <- colnames(CLASS_TOPHITS_signal)
	Anne_heatmap(log(heatmap_temp+1,2), experiment, "Heatmap of all Class genes log2(Signal)", "Heatmap_Class_Signals", "signal")

	
# 24b. Heatmap of Each Class Signal data
	logbook <- write_log(logbook, "# 24b. Heatmap of Each Class Signal data" )
	Class_GroupNames <- levels(my_class_genes$Group)
	filenames.ClassHeatmapsSignal <- list()
	filenames.Clusters <- list()
	for (i in 1:length(Class_GroupNames)) {
		Class_Genes <- subset(my_class_genes$GeneID, my_class_genes$Group %in% Class_GroupNames[i])
		Class_Group <- subset(mean_table_class, row.names(mean_table_class) %in% Class_Genes)
		if (dim(Class_Group)[1] > 1) {
			filename_prefix   <- paste("Heatmap_Class_Signal", Class_GroupNames[i], sep=".")
			filenames.ClassHeatmapsSignal <- c(filenames.ClassHeatmapsSignal, paste(experiment, "Heatmap_Class_Signal", Class_GroupNames[i] ,"png", sep="."))
			filenames.Clusters  <- c(filenames.Clusters, paste(experiment, filename_prefix ,"clusters", "txt", sep="."))
			main_title <- paste(experiment, "Heatmap of Class", Class_GroupNames[i], "log2(Signal)" , sep=" ")
			heatmap_temp <-(t(apply(Class_Group,1,as.numeric)))
			colnames(heatmap_temp) <- colnames(Class_Group)	
			Anne_heatmap(log(heatmap_temp+1,2), experiment, main_title, filename_prefix, "signal")
		}	
	}
	# save filenames
	filenames.ClassHeatmapsSignal
	filename <- paste(experiment, "ClassHeatmapsSignal.filenames.txt", sep=".")
	write.csv(filenames.ClassHeatmapsSignal, file=filename)
	#write.table(unlist(filenames.ClassHeatmapsSignal), file=filename, row.names = FALSE, col.names = FALSE, quote=FALSE)
	
	filename <- paste(experiment, "ClassHeatmapsSignal.clusters.filenames.txt", sep=".")
	temp <- NULL
	temp$files <- sapply(filenames.Clusters, paste0, collapse=",") 
	temp$names <- sapply(Class_GroupNames, paste0, collapse=",") 
	write.table(as.data.frame(temp), file=filename, row.names = FALSE, col.names = FALSE, quote=FALSE)


#--------------------------------------------------------------------- end heatmaps -------------------------------------------------------------------------
	
# 25. Signal data Expression plots	
	logbook <- write_log(logbook, "# 25. Signal data expression plots" )
	Class_signal_plots(log(CLASS_TOPHITS_signal,2),"Mean_signals_subClass_TOPHITS") 	


	
	
# 26-28. K-Means Clustering
	logbook <- write_log(logbook, "# 26. K-Means Clustering MedianFold" )
	if (length(row.names(TMEV_TOPHITS)) > 10) { KMeans_Clustering(TMEV_TOPHITS, "MedianFold", "log2Ratio") }
	logbook <- write_log(logbook, "# 27. K-Means Clustering Class Tophits" )
	if (length(row.names(CLASS_TOPHITS)) > 10) { KMeans_Clustering(CLASS_TOPHITS, "ClassTophits", "log2Ratio") }
	logbook <- write_log(logbook, "# 28. K-Means Clustering High Fold" )
	if (length(row.names(TMEV_TOPHITS_highfold)) > 10) { KMeans_Clustering(TMEV_TOPHITS_highfold, "HighFold", "log2Ratio") }
	if (length(row.names(CLASS_TOPHITS_signal)) > 10) { KMeans_Clustering(log(CLASS_TOPHITS_signal+0.1,2), "ClassTophitsSignal", "log2Signal") }

# 29 Signal K-means CLustering
	logbook <- write_log(logbook, "# 29. K-Means Clustering of Mean Signals" )
	# Clustering all genes is overkill, we will use only genes that changed at least weakly in one of the contrasts
	my_trimmed_mean_table <-  subset(mean_table, row.names(mean_table) %in% tophits_no_background)
	# take the mean table and remove rows with 1 zero or more
	my_trimmed_mean_table <- my_trimmed_mean_table[apply(my_trimmed_mean_table, 1, function(x) !any(x==0)),]
	head(my_trimmed_mean_table)
	if (length(row.names(mean_table)) > 10) { KMeans_Clustering(log(my_trimmed_mean_table,2), "MeanSignal", "log2Signal") }


# 30. Finally save the file container for downstream processing
	# Create menu
	menu1 <- NULL
	menu1$name <- 'Volcanoplots'
	menu1$items <- toJSON(items) 


	# Combine the lists in one object
	jSON$menus <- list()
	jSON$menus <- c(jSON$menus, toJSON(menu1))

# Finally convert the object to JSON and remove unwanted chars
	result <- toJSON(jSON)
	gsub("\\\\", "", result,fixed=TRUE)
	filename <- paste(experiment, "Container.json", sep=".")
	write.table(result, file=filename,  quote = FALSE, col.names = FALSE, , row.names = FALSE)