organism_helpers.rb in rbbt-sources-1.2.0

- old
+ new

@@ -1,5 +1,7 @@
+require 'net/ftp'
+
 $biomart_ensembl_gene = ['Ensembl Gene ID', 'ensembl_gene_id']
 $biomart_ensembl_protein = ['Ensembl Protein ID', 'ensembl_peptide_id']
 $biomart_ensembl_exon = ['Ensembl Exon ID', 'ensembl_exon_id']
 $biomart_ensembl_transcript = ['Ensembl Transcript ID', 'ensembl_transcript_id']
 
@@ -54,12 +56,14 @@
 $biomart_exon_phase = [
   $biomart_ensembl_transcript,
   ['Phase','phase'],
 ]
 
+$biomart_pfam= [
+  ["Pfam Domain", 'pfam'],
+]
 
-
 $biomart_exons = [
   $biomart_ensembl_gene,
   ['Exon Strand','strand'],
   ['Exon Chr Start','exon_chrom_start'],
   ['Exon Chr End','exon_chrom_end'],
@@ -69,10 +73,16 @@
 
 file 'scientific_name' do |t|
   File.open(t.name, 'w') do |f| f.write $scientific_name end
 end
 
+file 'ortholog_key' do |t|
+  raise "Ortholog key not defined. Set up $ortholog_key in the organism specific Rakefile; example $ortholog_key = 'human_ensembl_gene'" unless defined? $ortholog_key and not $ortholog_key.nil?
+
+  File.open(t.name, 'w') do |f| f.write $ortholog_key end
+end
+
 file 'identifiers' do |t|
   identifiers = BioMart.tsv($biomart_db, $biomart_ensembl_gene, $biomart_identifiers, [], nil, :namespace => $namespace)
   identifiers.unnamed =  true
 
   $biomart_identifiers.each do |name, key, prefix|
@@ -454,9 +464,52 @@
   goterms = BioMart.tsv($biomart_db, ['Chromosome Name', "chromosome_name"] , [] , [], nil, :type => :double, :namespace => $namespace)
 
   File.open(t.name, 'w') do |f| f.puts goterms end
 end
 
+rule /^chromosome_.*/ do |t|
+  chr = t.name.match(/chromosome_(.*)/)[1]
+
+  archive = File.basename(FileUtils.pwd) =~ /^([a-z]{3}[0-9]{4})$/i ? $1 : nil
+
+  release = case archive
+            when "may2009"
+              "release-54"
+            when "jun2011"
+              "release-64"
+            when nil
+              Open.read("http://www.ensembl.org/info/data/ftp/index.html", :nocache => true).match(/pub\/(\w+-\d+)\/fasta/)[1]
+            end
+
+
+  ftp = Net::FTP.new("ftp.ensembl.org")
+  ftp.login
+  ftp.chdir("pub/#{ release }/fasta/")
+  ftp.chdir($scientific_name.downcase.sub(" ",'_'))
+  ftp.chdir('dna')
+  file = ftp.nlst.select{|file| file =~ /chromosome\.#{ chr }\.fa/}.first
+
+  raise "Fasta file for chromosome not found: #{ chr } - #{ archive }, #{ release }" if file.nil?
+
+  Log.debug("Downloading chromosome sequence: #{ file }")
+  TmpFile.with_file do |tmpfile|
+    ftp.getbinaryfile(file, tmpfile)
+    Open.write(t.name, Open.read(tmpfile, :gzip => true).sub(/^>.*\n/,'').gsub(/\s/,''))
+    ftp.close
+  end
+end
+
+rule /^possible_ortholog_(.*)/ do |t|
+  other = t.name.match(/ortholog_(.*)/)[1]
+  other_key = Organism.ortholog_key(other).produce.read
+  BioMart.tsv($biomart_db, $biomart_ensembl_gene, [["Ortholog Ensembl Gene ID", "inter_paralog_" + other_key]], [], nil, :keep_empty => false, :type => :flat, :filename => t.name, :namespace => $namespace)
+end
+
+rule /^ortholog_(.*)/ do |t|
+  other = t.name.match(/ortholog_(.*)/)[1]
+  other_key = Organism.ortholog_key(other).produce.read
+  BioMart.tsv($biomart_db, $biomart_ensembl_gene, [["Ortholog Ensembl Gene ID", other_key]], [], nil, :keep_empty => false, :type => :flat, :filename => t.name, :namespace => $namespace)
+end
 
 rule /[a-z]{3}[0-9]{4}\/.*/i do |t|
   t.name =~ /([a-z]{3}[0-9]{4})\/(.*)/i
   archive = $1
   task    = $2