############################################################################################################
############################################################################################################
######################################### BioXM application ################################################
############################################################################################################
############################################################################################################
#                                                                                                          #
# Functional modules integrating essential cellular functions are predictive of the response of leukaemia  #
# cells to DNA damage                                                                                      #
#                                                                                                          #
# Supplementary Material                                                                                   #
# R-Code written by Nil Turan and Philipp Antczak                                                          #
#                                                                                                          #
# Dr. Francesco Falciani Group, University of Birmingham                                                   #
#                                                                                                          #
# 2008                                                                                                     #
#                                                                                                          #
############################################################################################################

############################################################################################################
############################################################################################################
# CALL FUNCTIONS ###########################################################################################
############################################################################################################
############################################################################################################

############################################################################################################
# (Others_BioXM-I) CALL FUNCTION: get adjacency list from a BioXM output
############################################################################################################
get.adj.list <- function( file.name, gene.source, gene.target, annotation, dataset.symbol ){

	tmp <- list()
	for( i in seq( file.name[,1] ) ) 
	{

          source <- strsplit( as.character( file.name[i,gene.source] )," \\[" )[[1]][1]
		target <- strsplit( as.character( file.name[i,gene.target] )," \\[" )[[1]][1]
		currenti <- as.character( file.name[i,annotation] )

		subtmp <- c( unlist( tmp[source] ),which( as.character(dataset.symbol) == target ) )
		subtmp2 <- c( unlist( tmp[target] ),which( as.character(dataset.symbol) == source ) )
		tmp[source] <- list( unique( subtmp ) )
		tmp[target] <- list( unique( subtmp2 ) )
		attributes( tmp[source] ) <- list( pathways=c( attributes( tmp[source] )$pathways,currenti ) )
	}

	return( tmp )
}


############################################################################################################
# (Others_BioXM-II) CALL FUNCTION: get a network from a BioXM output
############################################################################################################
get.network <- function( data,
			 bioxm.output.KEGG,
			 bioxm.output.pp, 
			 f.out.adj.list,
			 pp.col.target = "Gene.target",
			 pp.col.source = "Gene.source",
			 KEGG.col.target = "Target.Gene",
			 KEGG.col.source = "Source.Gene",
			 pp.annotation = "Annotation.Interaction.Source",
			 KEGG.annotation = "Pathway.source")
{
	pp.col.target <- gsub( " ","\\.", pp.col.target )
	pp.col.source <- gsub( " ","\\.", pp.col.source )
	KEGG.col.target <- gsub( " ","\\.", KEGG.col.target )
	KEGG.col.target <- gsub( " ","\\.", KEGG.col.source )
	pp.annotation <- gsub( " ","\\.", pp.annotation )
	KEGG.annotation <- gsub( " ","\\.", KEGG.annotation )

	# read the output created by BioXM
	kegg <- read.delim( bioxm.output.KEGG,header=T,row.names=NULL )
	pp <- read.delim( bioxm.output.pp,header=T,row.names=NULL )

	kegg.sub <- unique( kegg[, c( KEGG.col.source, KEGG.col.target, KEGG.annotation ) ] )

	if( class( data ) == "matrix" )
	{
		dataset.symbol <- rownames( data )
	}else if( class( data ) == "vector" )
	{
		dataset.symbol <- data
	}
  
	pp.sub <- unique( pp[, c( pp.col.source,pp.col.target,pp.annotation )] )

	# create KEGG interaction list
	stock.list <- c()
	for( i in kegg.sub[,KEGG.col.target] ){
		target.tmp <- unlist( strsplit( as.character( kegg.sub[i,KEGG.col.target] ),";" ) )
		source.tmp <- unlist( strsplit( as.character( kegg.sub[i,KEGG.col.source] ),";" ) )
		for( j in seq( target.tmp ) )
		{
			for( l in seq( source.tmp ) )
			{
				stock.list <- rbind( stock.list, c( source.tmp[l],target.tmp[j] ) )
			}
		}
	}

	kegg.sub <- cbind( stock.list, rep( unique( as.character( kegg.sub[,KEGG.annotation])), length( stock.list[,1] ) ) )
	colnames( kegg.sub ) <- c(KEGG.col.source, KEGG.col.target, KEGG.annotation)

	kegg.adj.list<-get.adj.list( kegg.sub, KEGG.col.source, KEGG.col.target, KEGG.annotation, dataset.symbol )
	list.order <- match( names( kegg.adj.list ), as.character( dataset.symbol ) )
	kegg.adj.list.2 <- list()
	length( kegg.adj.list.2 ) <- length( dataset.symbol )
	names( kegg.adj.list.2 ) <- as.character( dataset.symbol )

	for( i in seq( list.order ) ) kegg.adj.list.2[list.order[i]] <- list( kegg.adj.list[[i]] )
	kegg.adj.list <- kegg.adj.list.2
	rm( kegg.adj.list.2 )

	# create protein-protein interaction list
	pp.adj.list <- get.adj.list( pp.sub, pp.col.source, pp.col.target, pp.annotation, dataset.symbol )
	list.order <- match( names( pp.adj.list ), as.character( dataset.symbol ) )
	pp.adj.list.2 <- list()
	length( pp.adj.list.2 ) <- length( dataset.symbol )
	names( pp.adj.list.2 ) <- as.character( dataset.symbol )
	for( i in seq( list.order ) ) pp.adj.list.2[ list.order[i] ] <- list( pp.adj.list[[i]] )
	pp.adj.list <- pp.adj.list.2
	rm( pp.adj.list.2 )
   
	# merge the KEGG list and protein-protein interaction list
	super.adj.list <- list()
	super.adj.list[[1]] <- kegg.adj.list
	super.adj.list[[2]] <- pp.adj.list

	adj.list.merge <- list()
	length( adj.list.merge ) <- length( dataset.symbol )
	for( i in 1:length( dataset.symbol ) )
	{
		for( j in seq( super.adj.list ) )
		{
			adj.list.merge[i] <- list( unique( c( unlist( adj.list.merge[[i]] ), unlist( super.adj.list[[j]][[i]] ) ) ) )
		}
	}

	names( adj.list.merge ) <- dataset.symbol
	null.values <- which( sapply ( adj.list.merge, is.null ) )
	for( i in null.values ) adj.list.merge[[i]] <- integer(0)

	# write down adjacent list
	write.adj.list( adj.list.merge, f.out.adj.list )

	return( list( adj.list = adj.list.merge ) )

}