R/ParseGtf.r

In zhezhangsh/rchive:

# Load a GTF file to create a GRanges
ParseGtf<-function(fn.in, fn.out, path=paste(Sys.getenv("RCHIVE_HOME"), 'data/gene/public/gtf/r', sep='/'), Dbxref=TRUE, group=TRUE, separators=c()) {
  # fn.in, fn.out		Name of input/output files
  # path    			  Path to output file
  # Dbxref, group		Whether to further parse these attribute fields
  # separators  		The separators to be used to parse Dbxref or group field
  
  library(GenomicRanges);
  library(rtracklayer);
  
  if (!file.exists(path)) dir.create(path, recursive=TRUE);
  
  ##############################################################################
  ##############################################################################  
  # Function that performs the splitting
  splitColumn<-function(splt, separator) {
    n<-sapply(splt, length);
    pairs<-strsplit(unlist(splt, use.names=FALSE), separator);
    nm<-sapply(pairs, function(x) x[1]);
    cnm<-unique(nm);
    vl<-sapply(pairs, function(x) x[length(x)]);
    names(vl)<-rep(names(splt), n);
    mp<-split(vl, nm);
    df<-matrix('', nrow=length(gr), ncol=length(cnm), dimnames=list(names(gr), cnm));
    for (i in 1:length(cnm)) df[names(mp[[cnm[i]]]), cnm[i]]<-as.vector(mp[[cnm[i]]]); 
    df<-data.frame(df, stringsAsFactors=FALSE);
    colnames(df)<-tolower(colnames(df));
    df;
  }
  ##############################################################################
  ##############################################################################
  
  if (file.exists(paste(path, '/', fn.out, '_full.rds', sep=''))) {
    cat('Loading imported GTF in GRanges\n');
    gr<-readRDS(paste(path, '/', fn.out, '_full.rds', sep=''));
    meta<-elementMetadata(gr);
    rownames(meta)<-names(gr);
  } else {
    cat('Importing GTF file:', fn.in, '\n');
    gr<-rtracklayer::import(fn.in);
    cat('The loaded GTF file includes', length(gr), 'entries.\n');
    names(gr)<-1:length(gr);
    
    meta<-elementMetadata(gr);
    rownames(meta)<-names(gr);
    
    # Further parse 'Dbxref' column
    if (Dbxref & tolower('Dbxref') %in% tolower(colnames(meta))) {
      dbx<-BiocGenerics::as.vector(meta$Dbxref);
      if (length(separators) !=2) separators<-c('', ':');
      names(dbx)<-names(gr);
      df<-splitColumn(dbx, separators[2]);
      meta<-cbind(meta[, !(colnames(meta) %in% 'Dbxref')], df);
    }
    
    # Further parse 'group' column  
    if (group & 'group' %in% colnames(meta)) {
      grp<-as.vector(meta$group);
      if (length(separators) !=2) separators<-c('; ', ' ');
      splt<-strsplit(grp, separators[1]);
      names(splt)<-names(gr);
      df<-splitColumn(splt, separators[2]);
      meta<-cbind(meta[, !(colnames(meta) %in% 'group')], df);
    }
    
    elementMetadata(gr)<-meta;
    saveRDS(gr, file=paste(path, '/', fn.out, '_full.rds', sep=''));
  }
  ##############################################################################
  
  # Minimal metadata for slim version
  cnm<-strsplit("source;type;score;phase;id;transcript_id;transcript_name;transcript_type;gene_id;gene_name;gene_type;parent", ';')[[1]];
  cnm0<-tolower(colnames(meta));
  if(!('gene_id' %in% cnm0)) cnm0[cnm0=='geneid']<-'gene_id';
  colnames(meta)<-cnm0;
  meta<-meta[, colnames(meta) %in% cnm];
  cnm<-cnm[cnm %in% colnames(meta)];
  meta<-meta[, cnm];
  elementMetadata(gr)<-meta;
  saveRDS(gr, file=paste(path, '/', fn.out, '_slim.rds', sep=''));
  
  # split by type
  if ('type' %in% colnames(meta)) {
    tp<-as.vector(meta[['type']]);
    meta<-meta[, colnames(meta)!='type'];
    elementMetadata(gr)<-meta;
    gr0<-split(gr, tp);
    f<-sapply(names(gr0), function(nm) saveRDS(gr0[[nm]], file=paste(path, '/', fn.out, '_', tolower(nm), '.rds', sep='')));
  }
  
  paste(path, '/', fn.out, '_slim.rds', sep='');
}