#####
##
## Analyses H3K4me3
## 
## Marion Aguirrebengoa
## 
## Last modification : 10/17/2016
##
## R version 3.2.5 (2016-04-14)
##
#####

## REQUIREMENTS
library( "rtracklayer" );
library( "BSgenome.Hsapiens.UCSC.hg19" );
seqlens = seqlengths( Hsapiens );

## READ DATA
	# Random sites
rdm346 = import.bed( "Rdm_346_filtered_by_histone.bed" )
load( "rand.Rdata" );

	# Read AsiSI 80 best cleaved sites
asi80 = "BLESS_80best_JunFragPE_Rmdups_pm500bp.bed";
asi = list();
asi[["80"]] = import.bed( asi80 );

	# H3K4me3
bcbw = import.bw( "H3K4me3_BGI_RUN_3_bin50_log2ratio.bw",  as = "RleList" );
bw = list();
bw[["pOHT"]] = import.bw( "H3K4me3_OHT_normalized_hg19.bw", as = "RleList" );
bw[["mOHT"]] = import.bw( "H3K4me3_normalized_hg19.bw",  as = "RleList" );

	# Genes list
sh = read.table( "liste_best_299.bed", sep = "\t", header = TRUE );
str = sh[, "Strand"];
str[str == -1] = "-";
str[str == 1] = "+";
shgr = GRanges( seqnames = paste0( "chr", sh[, "Chromosome.Name"] ), strand = str, ranges = IRanges( start = sh[, "Gene.Start..bp."], end = sh[, "Gene.End..bp."] ), name = sh[, "Ensembl.Gene.ID"] );


## FUNCTIONS
	# Compute 1 value per site
compute1ValPerSite = function( bed, wig, w = 20000, seqlens, fun = "sum" ){
	if( class( wig ) != "SimpleRleList" ){
		stop( "ERROR : unknown class of wig, please provide a SimpleRleList" );
	}
	vec = NULL;
	for( i in 1:length( bed ) ){
		if( i %% 100 == 0 ){
			message( i, "/", length( bed ) );
		}
		bedi = bed[i, ];
		chr = as.character( seqnames( bedi ) );
		cov = wig[[chr]];
		center = start( bedi ) + 4;
		stW = center - w;
		edW = center + w;
		stW[stW < 1] = 1;
		edW[edW > seqlens[chr] | edW > length( cov )] = min( length( cov), seqlens[chr] );
		v = Views( cov, start = stW, end = edW );
		if( fun == "sum" ){
			vm = sum( v );
		}else if( fun == "mean" ){
			vm =  mean( v );
		}else{
			stop( "ERROR : unknown function fun = ", fun, " - fun must be 'sum' or 'mean'" );
		}
		vec = c( vec, vm );
	}
	return( vec );
}

	# Cut Rle in bin
resizeRle = function( rle, binSize ){
	if( is.null( rle ) ){
		return( rep( NA, binSize ) );
	}else{
		len = length( rle );
		step = len / binSize;
		st = floor( seq( 1, len, step ) );
		ed = floor( seq( step, len + step - 1, step ) );
		ed[ed > len] = len;
		v = Views( rle, start = st, end = ed );
		vm = viewApply( v, function( x ){
			return( mean( as.numeric( x ) ) )
		} );
		return( vm );
	}
}

	# Create a matrix from a list of Rle
computeMatrix = function( rleList, binSize ){
	nrleList = lapply( rleList, resizeRle, binSize = binSize );
	return( matrix( unlist( nrleList ), nrow = length( nrleList ), ncol = binSize, byrow = TRUE ) );
}

	# Get values on genes and intergenic zones
getGeneBody = function( coverage, bed, chrList = paste0( "chr", c( 1:22, "X" ) ), binSize = 100, w = 3000 ){
	rv = rvav = rvap = list();	
	sbed = bed[width( bed ) > binSize, ];
	if( length( sbed ) > 0 ){
		for( i in 1: length( sbed ) ){
			if( i %% 100 == 0 ){
				message( i, "/", length( sbed ) );
			}
			sbedi = sbed[i, ];
			str = as.character( strand( sbedi ) );
			chr = as.character( seqnames( sbedi ) );
			scov = coverage[[chr]];
			if( ( start( sbedi ) - w > 0 ) & ( end( sbedi ) + w < length( scov ) ) ){
				if( str == "+" ){
					vav = Views( scov, start = ( start( sbedi ) - w ), end = start( sbedi ) );
					v = Views( scov, start = start( sbedi ), end = end( sbedi ) );
					vap = Views( scov, start = end( sbedi ), end = ( end( sbedi ) + w ) );
					rvav[[i]] = vav[[1]];
					rv[[i]] = v[[1]];
					rvap[[i]] = vap[[1]];
				}else{
					vav = Views( scov, start = ( start( sbedi ) - w ), end = start( sbedi ) );
					v = Views( scov, start = start( sbedi ), end = end( sbedi ) );
					vap = Views( scov, start = end( sbedi ), end = ( end( sbedi ) + w ) );
					rvav[[i]] = rev( vap[[1]] );
					rv[[i]] = rev( v[[1]] );
					rvap[[i]] = rev( vav[[1]] );
				}
			}
		}
	}
	matav = computeMatrix( rvav, binSize = w / 100 );
	mat = computeMatrix( rv, binSize = binSize );
	matap = computeMatrix( rvap, binSize = w / 100 );
	matav = matav[rowSums( is.na( matav ) ) == 0, ];
	mat = mat[rowSums( is.na( mat ) ) == 0, ];
	matap = matap[rowSums( is.na( matap ) ) == 0, ];
	cb = cbind( matav, mat, matap );
	res = colMeans( cb );
	return( res );
}

## RUN BOXPLOT
	# Compute values
resAsiBC = list();
for( a in  names( asi ) ){
	message( a );
	resAsiBC[[a]] = compute1ValPerSite( bed = asi[[a]], wig = bcbw, w = 500, seqlens, fun = "mean" );
}
resRandBC = list();
for( r in  names( rand ) ){
	message( r );
	resRandBC[[r]] = compute1ValPerSite( bed = rand[[r]], wig = bcbw, w = 500, seqlens, fun = "mean" );
}
resAsi = list();
for( o in names( bw ) ){
	resAsi[[o]] = list();
	for( a in  names( asi ) ){
		message( o, " - ", a );
		resAsi[[o]][[a]] = compute1ValPerSite( bed = asi[[a]], wig = bw[[o]], w = 500, seqlens, fun = "mean" );
	}
}
resRand = list();
for( o in names( bw ) ){
	resRand[[o]] = list();
	for( r in  names( rand ) ){
		message( o, " - ", r );
		resRand[[o]][[r]] = compute1ValPerSite( bed = rand[[r]], wig = bw[[o]], w = 500, seqlens, fun = "mean" );
	}
}

	# Draw boxplot
postscript( "Boxplot_bamCompare_SitesAndRandoms.eps" );
par( mfrow = c( 1, 2 ) );
boxplot( resAsiBC[["80"]], resRandBC[["80"]], col = c( "dodgerblue3", "grey" ), main = "Boxplot_bamCompare_80SitesAndRandoms" );
axis( 1, 1:2, c( "AsiSIsites", "Random" ) );
dev.off();

	# Compute P-values
wilcox.test( resAsi[["mOHT"]][["80"]], resAsi[["pOHT"]][["80"]], paired = TRUE )$p.value; # 9.380713e-07
wilcox.test( resAsiBC[["80"]], mu = 0 )$p.value; # 4.425968e-05
wilcox.test( resRand[["mOHT"]][["80"]], resRand[["pOHT"]][["80"]], paired = TRUE )$p.value;
wilcox.test( resRandBC[["80"]], mu = 0 )$p.value;

## RUN PROFILE
	# Compute Profile on Shiloh genes
profsh = list();
for( o in names( bw ) ){
	message( o );
	profsh[[o]] = getGeneBody( coverage = bw[[o]], bed = shgr, binSize = 100, w = 3000 );
}

	# Draw all profiles
postscript( "Profiles.eps" );
plot( 1:length( profsh[["pOHT"]] ), profsh[["pOHT"]], type = "l", col = "red", ylim = c( 0, 4 ), xaxt = "n", xlab = "", ylab = "" )
points( 1:length( profsh[["mOHT"]] ), profsh[["mOHT"]], type = "l", col = "blue" )
title( "Shiloh" )
abline( v = 30, col = "grey" );
abline( v = 130, col = "grey" );
dev.off();