cistematic/genomes/ecaballus.py

   1 ###########################################################################
   2 #                                                                         #
   3 # C O P Y R I G H T   N O T I C E                                         #
   4 #  Copyright (c) 2003-10 by:                                              #
   5 #    * California Institute of Technology                                 #
   6 #                                                                         #
   7 #    All Rights Reserved.                                                 #
   8 #                                                                         #
   9 # Permission is hereby granted, free of charge, to any person             #
  10 # obtaining a copy of this software and associated documentation files    #
  11 # (the "Software"), to deal in the Software without restriction,          #
  12 # including without limitation the rights to use, copy, modify, merge,    #
  13 # publish, distribute, sublicense, and/or sell copies of the Software,    #
  14 # and to permit persons to whom the Software is furnished to do so,       #
  15 # subject to the following conditions:                                    #
  16 #                                                                         #
  17 # The above copyright notice and this permission notice shall be          #
  18 # included in all copies or substantial portions of the Software.         #
  19 #                                                                         #
  20 # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,         #
  21 # EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF      #
  22 # MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND                   #
  23 # NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS     #
  24 # BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN      #
  25 # ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN       #
  26 # CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE        #
  27 # SOFTWARE.                                                               #
  28 ###########################################################################
  29 #
  30 # data for Equus Caballus
  31 import string
  32 from cistematic.genomes import Genome
  33 from os import environ
  34
  35 if environ.get("CISTEMATIC_ROOT"):
  36     cisRoot = environ.get("CISTEMATIC_ROOT")
  37 else:
  38     cisRoot = "/proj/genome"
  39
  40 geneDB = "%s/E_caballus/ecaballus.genedb" % cisRoot
  41
  42
  43 def loadChromosome(db, chromID, chromPath, chromOut):
  44     seqArray = []
  45     ecGenome = Genome("ecaballus", dbFile=db)
  46     inFile = open(chromPath, "r")
  47     line = inFile.readline()
  48     for line in inFile:
  49         seqArray.append(line.strip())
  50
  51     seq = string.join(seqArray, "")
  52     seqLen = len(seq)
  53     if seqLen < 1:
  54         print "Problems reading sequence from file"
  55
  56     print "writing to file %s" % chromOut
  57     outFile = open("%s%s" % (cisRoot, chromOut), "w")
  58     outFile.write(seq)
  59     outFile.close()
  60     ecGenome.addChromosomeEntry(chromID, chromOut, "file")
  61
  62
  63 def loadGeneEntries(db, gFile):
  64     """ FIXME - NEED TO DEAL WITH ALTERNATIVE SPLICING ENTRIES
  65     """
  66     geneEntries = []
  67     alreadySeen = []
  68     ecGenome = Genome("ecaballus", dbFile=db)
  69     geneFile = open(gFile, "r")
  70     geneFile.readline()
  71     for line in geneFile:
  72         cols = line.split("\t")
  73         if cols[11].strip() != "GENE":
  74             continue
  75
  76         name = cols[10].split(":")
  77         gid = name[1]
  78         if gid == "" or gid in alreadySeen:
  79             continue
  80
  81         alreadySeen.append(gid)
  82         start = int(cols[2]) - 1
  83         stop = int(cols[3]) - 1
  84         sense = cols[4]
  85         chrom = cols[1].strip()
  86         if sense == "+":
  87             sense = "F"
  88         else:
  89             sense = "R"
  90
  91         geneID = ("ecaballus", gid)
  92         gidVersion = 1
  93         geneEntries.append((geneID, chrom, start, stop, sense, "gene", gidVersion))
  94
  95     print "Adding %d gene entries" % len(geneEntries)
  96     ecGenome.addGeneEntryBatch(geneEntries)
  97
  98
  99 def loadGeneFeatures(db, gFile):
 100     """ Load gene features such as CDS, UTR, and PSEUDO from the gene file.
 101     """
 102     featureEntries = []
 103     ecGenome = Genome("ecaballus", dbFile=db)
 104     featureFile = open(gFile, "r")
 105     featureFile.readline()
 106     for line in featureFile:
 107         cols = line.split("\t")
 108         if cols[11].strip() not in ["CDS", "UTR", "PSEUDO"]:
 109             continue
 110
 111         fType = cols[11]
 112         name = cols[10].split(":")
 113         gid = name[1]
 114         if gid == "":
 115             continue
 116
 117         start = int(cols[2]) - 1
 118         stop = int(cols[3]) - 1
 119         sense = cols[4]
 120         chrom = cols[1].strip()
 121         if sense == "+":
 122             sense = "F"
 123         else:
 124             sense = "R"
 125
 126         geneID = ("ecaballus", gid)
 127         gidVersion = 1
 128         featureEntries.append((geneID, gidVersion, chrom, start, stop, sense, fType))
 129
 130     print "Adding %d feature entries" % len(featureEntries)
 131     ecGenome.addFeatureEntryBatch(featureEntries)
 132
 133
 134 def createDBFile(db):
 135     ecGenome = Genome("ecaballus",  dbFile=db)
 136     ecGenome.createGeneDB(db)
 137
 138
 139 def createDBindices(db):
 140     ecGenome = Genome("ecaballus", dbFile=db)
 141     ecGenome.createIndices()
 142
 143
 144 def buildHorseDB(db=geneDB):
 145     genePath = "%s/download/seq_gene.md" % cisRoot
 146     chromos = {"1": "%s/download/chr1.fa" % cisRoot,
 147                "2": "%s/download/chr2.fa" % cisRoot,
 148                "3": "%s/download/chr3.fa" % cisRoot,
 149                "4": "%s/download/chr4.fa" % cisRoot,
 150                "5": "%s/download/chr5.fa" % cisRoot,
 151                "6": "%s/download/chr6.fa" % cisRoot,
 152                "7": "%s/download/chr7.fa" % cisRoot,
 153                "8": "%s/download/chr8.fa" % cisRoot,
 154                "9": "%s/download/chr9.fa" % cisRoot,
 155                "10": "%s/download/chr10.fa" % cisRoot,
 156                "11": "%s/download/chr11.fa" % cisRoot,
 157                "12": "%s/download/chr12.fa" % cisRoot,
 158                "13": "%s/download/chr13.fa" % cisRoot,
 159                "14": "%s/download/chr14.fa" % cisRoot,
 160                "15": "%s/download/chr15.fa" % cisRoot,
 161                "16": "%s/download/chr16.fa" % cisRoot,
 162                "17": "%s/download/chr17.fa" % cisRoot,
 163                "18": "%s/download/chr18.fa" % cisRoot,
 164                "19": "%s/download/chr19.fa" % cisRoot,
 165                "20": "%s/download/chr20.fa" % cisRoot,
 166                "21": "%s/download/chr21.fa" % cisRoot,
 167                "22": "%s/download/chr22.fa" % cisRoot,
 168                "23": "%s/download/chr23.fa" % cisRoot,
 169                "24": "%s/download/chr24.fa" % cisRoot,
 170                "25": "%s/download/chr25.fa" % cisRoot,
 171                "26": "%s/download/chr26.fa" % cisRoot,
 172                "27": "%s/download/chr27.fa" % cisRoot,
 173                "28": "%s/download/chr28.fa" % cisRoot,
 174                "29": "%s/download/chr29.fa" % cisRoot,
 175                "30": "%s/download/chr30.fa" % cisRoot,
 176                "31": "%s/download/chr31.fa" % cisRoot,
 177                "M": "%s/download/chrM.fa" % cisRoot,
 178                "X": "%s/download/chrX.fa" % cisRoot,
 179                "Un": "%s/download/chrUn.fa" % cisRoot
 180     }
 181
 182     print "Creating database %s" % db
 183     createDBFile(db)
 184
 185     print "Adding gene entries"
 186     loadGeneEntries(db, genePath)
 187
 188     print "Adding gene features"
 189     loadGeneFeatures(db, genePath)
 190
 191     for chromID in chromos.keys():
 192         print "Loading chromosome %s" % chromID
 193         loadChromosome(db, chromID, chromos[chromID], "/E_caballus/chromo%s.bin" % chromID)
 194
 195     print "Creating Indices"
 196     createDBindices(db)
 197
 198     print "Finished creating database %s" % db