R/networkPlot.R

In HTSanalyzeR: Gene set over-representation, enrichment and network analyses for high-throughput screens

##This function takes in a subnetwork module resulted from function
##networkAnalysis, a vector of labels for nodes in the module and a
##phenotype vector (optional) and generate a figure stored to
##'filepath' with the name 'filename'.

networkPlot <- function(nwAnalysisOutput, phenotypeVector = NULL) {
	##check arguments
	if(!is.list(nwAnalysisOutput) || 
		!(c("subnw") %in% names(nwAnalysisOutput)))
		stop("'nwAnalysisOutput' should contain a subnetwork module!\n")
	subnw <- nwAnalysisOutput$subnw
	labels <- nwAnalysisOutput$labels
	if(!is(subnw, "graphNEL"))
		stop("The module in 'nwAnalysisOutput' should be an object ",
			"of class 'graphNEL'!\n")
	##If no phenotype vector is specified, then we can just plot the module	
	if(is.null(phenotypeVector)) {
		##png("EnrichedSubNw.png", width = 900, height = 900)
		plotModule(subnw, labels = labels)
		##dev.off()
	} else {
		paraCheck("phenotypes", phenotypeVector)
		##'diff.expr' holds the phenotype for the nodes of the sub-network	
		diff.expr <- phenotypeVector[nodes(subnw)]
		names(diff.expr) <- nodes(subnw)
		##'present' contains the information of wether a node has an 
		##associated phenotype (1) or not (-1), will be used to give a 
		##different shape to the nodes of the network
		present <- rep(1,length(nodes(subnw)))
		present[which(is.na(diff.expr))] <- -1
		##replaces all phenotypes of non-phenotyped nodes by a zero		
		diff.expr[which(is.na(diff.expr))] <- 0
		names(present) <- nodes(subnw)
		##Plot the module	
		if(length(nodes(subnw)) == 1) {
			##png(file.path(filepath, filename), width = 900, height = 900)
			plotModule(subnw, labels = labels, scores = present, 
				diff.expr = diff.expr)
			##dev.off()
		} else {
			Tcolors <- diff.expr
			Tcolors2 <- diff.expr
			if(max(abs(Tcolors)) < 5)  
				Tcolors <- Tcolors*5
			## set red colors
			if(any(Tcolors > 0)){
				maxRed <- max(ceiling(abs(Tcolors[which(Tcolors > 0)])))
				redCols <- colorRampPalette(colors = c("white", "red"))
				redVec <- redCols(maxRed)
				Tcolors2[which(Tcolors > 0)] <- 
					redVec[ceiling(abs(Tcolors[which(Tcolors > 0)]))]
			}
			##set the greens
			if(any(Tcolors < 0)){
				maxGreen <- max(ceiling(abs(Tcolors[which(Tcolors < 0)])))
				greenCols <- colorRampPalette(colors = c("white", "green"))
				greenVec <- greenCols(maxGreen)
				Tcolors2[which(Tcolors < 0)] <- 
					greenVec[ceiling(abs(Tcolors[which(Tcolors < 0)]))]
			}	  	
			colScale <- unique(Tcolors2)
			colboundary <- rep(0, length(colScale))
			colboundary <- sapply(colScale, function(c) {
				values <- diff.expr[which(Tcolors2 == c)]
				values[which(abs(values) == max(abs(values)))[1]]			
			})

			colMatrix <- cbind(colboundary[order(colboundary)], 
				colScale[order(colboundary)])
			##png(file.path(filepath, filename), width = 900, height = 900)
			plotModule(subnw, labels=labels, scores=present, 
				diff.expr=diff.expr)
			points(
				x = rep(-1.2, length(unique(Tcolors2))), 
				y = seq(1.2, (1.2-(0.05*length(colMatrix[, 2]))), 
				length.out = length(colMatrix[, 2])), 
				pch = 15, col = colMatrix[, 2]
			)
			text(
				x = rep(-1.1, length(unique(Tcolors2))), 
				y = seq(1.2, (1.2-(0.05*length(colMatrix[, 2]))), 
				length.out = length(colMatrix[, 2])),
				labels = signif(as.numeric(colMatrix[, 1]), digits = 2), 
				cex = 0.8
			)
			##dev.off()
		}
	}
}