comparison mtls_analyze/map_ChIP_peaks_to_genes.R @ 4:b465306d00ba draft default tip

Uploaded

4:b465306d00ba

##  .-.-.   .-.-.   .-.-.   .-.-.   .-.-.   .-.-.   .-.-.   .-.-.
## /|/ \|\ /|/ \|\ /|/ \|\ /|/ \|\ /|/ \|\ /|/ \|\ / / \ \ / / \ \
##`-'   `-`-'   `-`-'   `-`-'   `-`-'   `-`-'   `-`-'   `-`-'   ' '
## Nov 2011 th17
## Bonneau lab - "Aviv Madar" <[e-mail]>, 
## NYU - Center for Genomics and Systems Biology
##  .-.-.   .-.-.   .-.-.   .-.-.   .-.-.   .-.-.   .-.-.   .-.-.
## /|/ \|\ /|/ \|\ /|/ \|\ /|/ \|\ /|/ \|\ /|/ \|\ / / \ \ / / \ \
##`-'   `-`-'   `-`-'   `-`-'   `-`-'   `-`-'   `-`-'   `-`-'   ' '

rm(list=ls())
debug=F
script.version=0.1
print.error <- function(){
  cat("
NAME
       map_ChIP_peaks_to_genes.R

DESCRIPTION
       This script takes a MACS tab delimited file as input and
       produces two files:

       1) genes.xls
          A gene centered table that stores the peaks that are
          proximal (within x[kb] of TSS) and distal (within the
          gene's region + y[kb] upstream or downstream of TSS and
          TES respectively).

       2) peaks.xls
          A peak centered table that equals the MACS input table plus
          colums for proximal and distal targets of each peak and
          their distance from TSS (if exist). For gene.xls script also
          gives a poisson model based -log10(pval) score for proximal
          and genewide enrichment of peaks for each gene with at least
          one prox/dist peak associated with it.

ARGUMENTS
       input_file=path
         Path to MACS file.

       refseq_table=path
         Path to refseq table (gives TSS/TES locations for all genes).

       path_output=path
         Path to save genes.xls and peaks.xls output files.

       tss.dist=N
         The absolute distance from TSS where we connect a peak to
         gene (for proximal peaks).

       gene.wide.dist=N
         The absolute distance from TSS or TES where we connect a peak
         to gene (for peaks hitting anywhere in gene).

       effective.genome.size=N
         The effective mappable genome size (for mm9 the value is
         1.87e9. For hg19 the value is 2.7e9 (from MACS manual)).

       macs.skip.lines=N
         Number of lines to skip in input_file.

EXAMPLE 
       Rscript map_ChIP_peaks_to_genes_v.1.R \\
         input_file=SL971_SL970_peaks.xls \\
         refseq_table=UCSC_mm9_refseq_genes_Sep_15_2011.txt \\
         path_output=./ \\
         tss.dist=5000 \\
         gene.wide.dist=10000 \\
         effective.genome.size=1.87e9 \\
         macs.skip.lines=23

CITATION
       Please cite us if you used this script: The transcription
       factor network regulating Th17 lineage specification and
       function.  Maria Ciofani, Aviv Madar, Carolina Galan, Kieran
       Mace, Ashish Agarwal, Kim Newberry, Richard M. Myers, Richard
       Bonneau and Dan R. Littman et. al. (in preperation).

")
}

# retrieve args
if(debug==T){
	cmd.args <- c(
		# "input_file=input/th17/used_for_paper/rawData/MACS_Sep_15_2011/SL971_SL970_peaks.xls",
		"input_file=input/th17/used_for_paper/rawData/MACS_Sep_15_2011/SL3594_SL3592_peaks.xls",		
		"refseq_table=input/th17/used_for_paper/rawData/UCSC_mm9_refseq_genes_Sep_15_2011.txt",
		"path_output=/Users/aviv/Desktop/test_script/",
		"tss.dist=5000",
		"tes.dist=10000",
		"effective.genome.size=1.87e9",
		"gene.wide.dist=10000",
		"macs.skip.lines=23"
	)
} else {
	cmd.args <- commandArgs();      
}

if(length(grep("--version",cmd.args))){
	cat("version",script.version,"\n")
	q()
}

arg.names.cmd.line <- sapply(strsplit(cmd.args,"="),function(i) i[1])
args.val.cmd.line <- sapply(strsplit(cmd.args,"="),function(i) i[2])

arg.nms <- c("input_file","refseq_table","path_output","tss.dist",
			 "gene.wide.dist","effective.genome.size","macs.skip.lines")
arg.val <- character(length=length(arg.nms))
for(i in 1:length(arg.nms)){
	ix <- which(arg.names.cmd.line==arg.nms[i])
	if(length(ix)==1){
		arg.val[i] <- args.val.cmd.line[ix]
	} else {
		stop("######could not find ",arg.nms[i]," arg######\n\n",print.error())
		
	} 
}
if(debug==T){
	print(paste(arg.nms,"=",arg.val))
}
# the files here adhere to tab delim format
path.input.macs <- arg.val[1]
path.input.refseq <- arg.val[2]
path.output <- arg.val[3]
tss.dist <- as.numeric(arg.val[4])
gene.wide.dist <- as.numeric(arg.val[5])
effective.genome.size <- as.numeric(arg.val[6])
macs.skip.lines=as.numeric(arg.val[7])

if(debug==T){
# if(1){
	load("/Users/aviv/Desktop/r.u.RData")
	gn.nms <- rownames(r.u)
} else {
	##################
	### step 1 handle refseq table file (read, make unique)
	cat("reading refseq table",path.input.refseq,"\n")
	refseq <- read.delim(file=path.input.refseq)
	# refseq can have many transcripts for each gene
	# here i make it have only one transcript for each gene (the longest one)
	cat("for transcripts matching more than one gene keeping only the longest transcript\n")
	refseq$name2 <- as.character(refseq$name2)
	refseq$chrom <- as.character(refseq$chrom)
	refseq$strand <- as.character(refseq$strand)
	gn.nms <- unique(refseq$name2)
	#x <- sort(table(refseq$name2),decreasing=T)
	#gn.nms.non.unique <- names(x)[which(x>1)]
	#gn.nms.unique <- names(x)[which(x==1)]
	# create refseq unique r.u
	n <- length(gn.nms)
	r.u <- data.frame(cbind(chrom=rep("NA",n),strand=rep("NA",n),txStart=rep(0,n),txEnd=rep(0,n)),stringsAsFactors=FALSE,row.names=gn.nms)
	for(i in 1:n){
	  ix <- which(refseq$name2==gn.nms[i])
	  if(length(ix)==0) {
	    error("could not find gene", ng.nms[i], "in refseq table.  Bailing out...\n")
	  } else if (length(ix)>1){
	    l <- apply(refseq[ix,c("txStart","txEnd")],1,function(i) abs(i[1]-i[2]) )
	    l.max <- max(l)
	    ix <- ix[which(l==l.max)[1]]
	  }
	  r.u[gn.nms[i],c("chrom","strand","txStart","txEnd")] <- refseq[ix,c("chrom","strand","txStart","txEnd")]
	}
	r.u[,"txStart"] <- as.numeric(r.u[,"txStart"])
	r.u[,"txEnd"] <- as.numeric(r.u[,"txEnd"])

	# switch TSS and TES if we have chr "-"
	cat("correcting TSS, TES assiments in refseq table based on strand\n")
	ix <- which(r.u$strand=="-")
	tmp.tes <- r.u$txStart[ix]
	tmp.tss <- r.u$txEnd[ix]
	r.u[ix,c("txStart","txEnd")] <- cbind(tmp.tss,tmp.tes)
	# ############
}


#### step 2 read macs file
cat("reading Macs file",path.input.macs,"\n")

# read macs file
M <- read.delim(file=path.input.macs,skip=macs.skip.lines)
trgt.prox <- NA
trgt.dist <- NA
dtss.prox <- NA
dtss.dist <- NA
M.annot <- cbind(M,trgt.prox,trgt.dist,dtss.prox,dtss.dist)
M.annot[is.na(M.annot)] <- ""
# get the number of genes
m <- length(gn.nms)
if(debug==T){
	# m <- 100
}
f.nm.peak <- paste(sep="",path.output,"peaks.xls")
f.nm.gene <- paste(sep="",path.output,"genes.xls")
# for each genes in the expt (j goes over genes)
cat(file=f.nm.gene,sep="\t","Gene_ID","prox_n_peak","genewide_n_peak","gn_length","strand","prox_mean_pval","genewide_mean_pval",
			"prox_max_pval","genewide_max_pval",
			"prox_pois_model_pval","genewide_pois_model_pval","(peak_id,summit,d_TSS,d_TES,class,pval,fold_enrich,FDR),(..)\n")
peaks.summit <- (M[,"start"]+M[,"summit"])
cat(sep="","mapping MACS peaks to genes\n")
for (j in 1:m){
  if(j%%20==0){cat(".")}
  tss <- r.u[j,"txStart"]
  tes <- r.u[j,"txEnd"]
  gn.lngth <- abs(tss-tes)
  strand <- r.u[j,"strand"]
  chr <- r.u[j,"chrom"]
  if(strand=="+"){
    d.tss.all <- peaks.summit-tss
   	d.tes.all <- peaks.summit-tes
  } else {
    d.tss.all <- tss-peaks.summit
   	d.tes.all <- tes-peaks.summit
  }
  ix.distal <- sort(union( c(which( abs(d.tss.all) < gene.wide.dist ),which( abs(d.tes.all) < gene.wide.dist )),
               which( sign(d.tss.all) != sign(d.tes.all) )),decreasing=TRUE)
  ix.prox <- sort(which( abs(d.tss.all) < tss.dist ),decreasing=TRUE)	
  ix.prox <- ix.prox[which(M[ix.prox,"chr"]==chr)]    
  ix.distal <- ix.distal[which(M[ix.distal,"chr"]==chr)]
  n.peaks.prox <- length(ix.prox)
  n.peaks.distal <- length(ix.distal)
	# if there is at least one peak hitting gene j
  if(n.peaks.distal > 0){
    # for each peak (l goes over peaks)
    peaks.line <- paste(sep="","(")
    for (k in 1:n.peaks.distal){
      if(k>1){
        peaks.line <- paste(sep="",peaks.line,";(")
      }
      d.tss <- d.tss.all[ ix.distal[k] ]
      d.tes <- d.tes.all[ ix.distal[k] ]
      if(sign(d.tss)!=sign(d.tes)){
        class="intra"
      } else if(d.tss<=0){
        class="upstream"
      } else if(d.tss>0){
        class="downstream"
      }
      peaks.line <- paste(sep="",peaks.line,paste(sep=",", ix.distal[k], peaks.summit[ ix.distal[k] ], 
					d.tss, d.tes, class, M[ ix.distal[k] ,7], M[ ix.distal[k] ,8], M[ ix.distal[k] ,9] ),")")
	  # handle M.annot
	  dtss <- d.tss.all[ ix.distal[k] ]
	  prev.trgt <- M.annot[ix.distal[k],"trgt.dist"]
	  if(prev.trgt!=""){ # if peak already is with trgt choose one with min dtss
		if(abs(dtss)<abs(as.numeric(M.annot[ix.distal[k],"dtss.dist"]))){
			M.annot[ix.distal[k],"dtss.dist"] <- dtss
			M.annot[ix.distal[k],"trgt.dist"] <- gn.nms[j]
		}
	  } else { # if peak does not have trgt add current trgt
		M.annot[ix.distal[k],"dtss.dist"] <- dtss
		M.annot[ix.distal[k],"trgt.dist"] <- gn.nms[j]
	  }
	# calc the pval for these many peaks in proximal region and in gene-wide region
	}
	if(n.peaks.prox > 0){
		mean.pval.prox <- mean(M[ ix.prox ,7])
		max.pval.prox <- max(M[ ix.prox ,7])
		expected.num.peaks.genome.wide.prox <- length(which(M[,7]>=mean.pval.prox))
		# lambda = num peaks / genome size * searched region
		lambda.prox <- expected.num.peaks.genome.wide.prox/effective.genome.size*(tss.dist*2)		
		pval.pois.prox <- -log10(ppois(n.peaks.prox,lambda.prox,lower.tail=FALSE))
		# handle M.annot
		for(k in 1:n.peaks.prox){
			dtss <- d.tss.all[ ix.prox[k] ]
			prev.trgt <- M.annot[ix.prox[k],"trgt.prox"]
			if(prev.trgt!=""){ # if peak already is with trgt choose one with min dtss
				if(abs(dtss)<abs(as.numeric(M.annot[ix.prox[k],"dtss.prox"]))){
					M.annot[ix.prox[k],"dtss.prox"] <- dtss
					M.annot[ix.prox[k],"trgt.prox"] <- gn.nms[j]
				}
			} else { # if peak does not have trgt add current trgt
				M.annot[ix.prox[k],"dtss.prox"] <- dtss
				M.annot[ix.prox[k],"trgt.prox"] <- gn.nms[j]
			}
		}
	} else {
		mean.pval.prox <- "NA"
		max.pval.prox <- "NA"
		pval.pois.prox <- "NA"
	}
	mean.pval.distal <- mean(M[ ix.distal ,7])
	max.pval.distal <- max(M[ ix.distal ,7])
	expected.num.peaks.genome.wide.distal <- length(which(M[,7]>=mean.pval.distal))
	# lambda = num peaks / genome size * searched region
	lambda.distal <- expected.num.peaks.genome.wide.distal/effective.genome.size*(gn.lngth+gene.wide.dist*2)
    pval.pois.distal <- -log10(ppois(n.peaks.distal,lambda.distal,lower.tail=FALSE))
    # pring peaks for gene j    
    core.line <- paste(sep="\t",gn.nms[ j ],n.peaks.prox,n.peaks.distal,gn.lngth,strand,
							mean.pval.prox,mean.pval.distal,max.pval.prox,max.pval.distal,
							pval.pois.prox,pval.pois.distal)
    cat(file=f.nm.gene,append=TRUE,sep="",core.line,"\t",peaks.line,"\n")
  }
}
cat("Done!\n")

cat(file=f.nm.peak,colnames(M.annot),"\n",sep="\t")
write.table(file=f.nm.peak,append=T,col.names=F,row.names=F,M.annot,sep="\t")