R/getlength.R

In aiminy/goseq_test: Gene Ontology analyser for RNA-seq and other length biased data

#############################################################################
#Description: Fetches gene length data for the genome and id specified.
#Notes:
#Author: Matthew Young, Nadia Davidson
#Date Modified: 7/4/2015

getlength=function(genes,genome,id){
	#geneLenDataBase contains all the gene length data, check it is installed
	if(!any(library()$results[,"Package"]=="geneLenDataBase")){
		stop("Package geneLenDataBase is required for automatic gene length lookup.")
	}
	#Get a list of the length data files that are in the package
	repo=grep(".*\\..*\\.LENGTH",as.data.frame(data(package="geneLenDataBase")$results,
           stringsAsFactors=FALSE)$Item,ignore.case=TRUE,value=TRUE)
	#Is this genome/id in the local repository?
	if(paste(genome,id,"LENGTH",sep='.')%in%repo){
	   #Yes it is, so load it and group transcript lengths into genes
	   message("Loading ",genome," length data...")
	   data(list=paste(genome,id,"LENGTH",sep='.'),package='geneLenDataBase')
	   data=split(get(paste(genome,id,"LENGTH",sep='.'))$Length,get(paste(genome,id,"LENGTH",sep='.'))$Gene)

	   #Get the names of the genes used in the genome package
	   gene_names=get(paste(genome,id,"LENGTH",sep="."))$Gene

	} else { ### Try to find the gene lengths from TXDB annotation packages
	   message("Can't find ",genome,"/",id," length data in genLenDataBase...")

	   ## Is there a TxDB package installed for the organism?
	   installedPackages=rownames(installed.packages())

	   #first try looking for a match with the same type of gene IDs
	   txdbPattern=paste("TxDb","*",genome,id,sep=".")
	   txdbPack=installedPackages[grep(txdbPattern,installedPackages)]

	   #if not look for the package with generic gene IDs and convert later
	   if(length(txdbPack)==0){ 
	      orgstring=as.character(.ORG_PACKAGES[match(gsub("[0-9]+",'',genome),names(.ORG_PACKAGES))])
	      packageExtension=tail(strsplit(orgstring,".",fixed=T)[[1]],n=1)
	      genericID=names(.ID_MAP[which(.ID_MAP==packageExtension)])
	      txdbPattern=paste("TxDb","*",genome,genericID,sep=".")  
	      txdbPack=installedPackages[grep(txdbPattern,installedPackages)]
	      if(length(txdbPack)>0){
	      	 library(paste(orgstring,"db",sep='.'),character.only=TRUE)
	      	 userid=as.character(.ID_MAP[match(id,names(.ID_MAP))])
	      	 user2core=toTable(get(paste(orgstring,userid,sep='')))
	   	}
	   }
	   if(length(txdbPack)>0){
	      message("Found the annotaion package, ",txdbPack)
	      message("Trying to get the gene lengths from it.")
	      library(txdbPack,character.only=TRUE)
	      txdb<-get(txdbPack)
	      txlens<-transcriptLengths(txdb)
	      core2len=split(txlens$tx_len,txlens$gene_id)
	      if(length(grep(id,txdbPattern))!=0){
	         data=core2len ; gene_names=names(data)
	      } else { #We need to conver between gene IDs. Just use Matt's code from getgo.R
                 #Now we need to replicate the core IDs that need replicating
                 core2len=core2len[match(user2core[,1],names(core2len))]
                 #Now we can replace the labels on this list with the user ones from user2core,
                 #but there will be duplicates, so we have to unlist, label, then relist
                 data=split(unlist(core2len,FALSE,FALSE),rep(user2core[,2],sapply(core2len,length)))
	         gene_names=names(data)
	     }
	   } else { ## Can't use TxDB
	      if(length(txdbPack)==0 & any(.TXDB_ORGS==genome)) {
	         message("A TxDb annotation package exists for ",genome,". Consider installing it to get the gene lengths.",sep="")
	      }
	      #If it's not in geneLenDataBase, try to load the data from UCSC using rtracklayer
	      if(!require("rtracklayer")){
		  stop("Can't load the rtracklayer library. Is it installed?...")
	      }
	      message("Trying to download from UCSC. ",
			 "This might take a couple of minutes. ")

	      tryCatch( {
			 
		#If not, then the user will have to supply it themselves
		#download the relevant table from UCSC
		session <- browserSession()
		genome(session)<-genome
		if(id=="geneSymbol") #take the geneSymbols from refSeq genes
		   query <- ucscTableQuery(session,"refGene")
		else 
		   query <- ucscTableQuery(session,id)
		transTable<-getTable(query)

		# all this below is to get the length information of each gene
		starts=strsplit(as.character(transTable$exonStart),",")
		ends=strsplit(as.character(transTable$exonEnd),",")
		exons=mapply(data.frame, starts, ends, SIMPLIFY=FALSE)
		get_length<-function(x){ 
		   sum(as.integer(as.character(x[,2]))-as.integer(as.character(x[,1])) + 1 )  
                }
		lengths=sapply(exons,get_length)

		# now we need to get the correct list of gene names.
		names2=as.character(transTable$name2)
		gene_names=names2
		if(!any(toupper(genes)%in%toupper(names2))){
		   #check which column has the gene names
		   names1=as.character(transTable$name1)   
		   if(any(toupper(genes)%in%toupper(names1))){
			gene_names=names1
		   }
		}
		data<-split(lengths,gene_names)

                }, 
		error=function(e){ 
		   stop("Length information for genome ",genome," and gene ID ",id,
			" is not available. You will have to specify bias.data manually.") 
		}
            	)
	    }
	}
	#In order to assign just one length to each gene, we store the Median, Min Max 
	#and count of transcripts for each gene for later use
	len=data.frame(Gene=names(data),Median=sapply(data,median),Min=sapply(data,min),
			Max=sapply(data,max),Count=sapply(data,length))

	#How many of the genes match the names the user specified with "genes"
	matched_frac=sum(toupper(genes)%in%toupper(gene_names))/length(genes)
	#None? Throw an error.
	if(matched_frac==0){
		stop("The gene names specified do not match the gene names for genome ",
		genome," and ID ",id,".\n\tGene names given were: ",paste(genes[1:10],collapse=", "),
		"\n\tRequired gene names are: ",paste(gene_names[1:10],collapse=", "))
	}
	#Less than 60%? Give a warning and continue
	if(matched_frac<.6){
		warning("More than 40% of gene names specified did not match the gene names for genome ",
		genome," and ID ",id,".  No length data will be available for these genes.\n\tGene names which failed to match were: ",
		paste(genes[!genes%in%gene_names][1:10],collapse=", "),"\n\tRequired gene names are: ",
		paste(gene_names[1:10],collapse=", "))
	}
	#Finally, return the median transcript length for each matching gene
	return(len$Median[match(toupper(genes),toupper(len$Gene))])
}