R/generateNetworkTable.R

In hafenr/SECprofiler: Complex feature detection using SEC protein chromatograms.

#' From protein level Traces, generate a Network table with pairwise profile
#' correlation score
#' 
#' @param Traces A Traces object containing wide format elution profiles (protein.traces).
#' @param correlation_cutoff Minimum pearson correlation value to report an edge between two proteins.
#' @param write_file Logical (TRUE/FALSE) whether an output file named [systime]_NetworkTable.csv shall be written to the working folder.
#' @return A data.table containing columns UniProt_A UniProt_B and pearson_correlation, readable e.g. by Cytoscape.
#' @export

generateNetworkTable <- function(Traces, correlation_cutoff = 0.9, write_csv = TRUE) {
	data <- Traces$traces
	# calculate protein level correlation matrix
	data.traces <- as.matrix(data[,2:ncol(data)])
	rownames(data.traces) <- data$protein_id
	corrmtrx<-cor(t(data.traces))
	corrlist<-corrmtrx
	corrlist[lower.tri(corrlist,diag=TRUE)]=NA
	corrlist=as.data.frame(as.table(corrlist))
	corrlist=na.omit(corrlist)
	corrlist=corrlist[order(-abs(corrlist$Freq)),]
	names(corrlist)<-c("Uniprot_A", "Uniprot_B", "pearson_correlation")
	corrlist.s <- subset(corrlist, pearson_correlation >= correlation_cutoff)
	if (write_csv == TRUE) {
		write.csv(corrlist.s, file=paste0(deparse(substitute(Traces)), "_NetworkTable.csv", quote = FALSE, row.names = FALSE))
	}
	return(corrlist.s)
}