cistematic/genomes/ggallus.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 Gallus gallus
  31 import string
  32 from cistematic.genomes import Genome
  33 from cistematic.core.geneinfo import geneinfoDB
  34 from os import environ
  35
  36 if environ.get("CISTEMATIC_ROOT"):
  37     cisRoot = environ.get("CISTEMATIC_ROOT")
  38 else:
  39     cisRoot = "/proj/genome"
  40
  41 geneDB = "%s/G_gallus/ggallus.genedb" % cisRoot
  42
  43
  44 def loadChromosome(db, chromID, chromPath, chromOut):
  45     seqArray = []
  46     ggGenome = Genome("ggallus", dbFile=db)
  47     inFile = open(chromPath, "r")
  48     line = inFile.readline()
  49     for line in inFile:
  50         seqArray.append(line.strip())
  51
  52     seq = string.join(seqArray, "")
  53     seqLen = len(seq)
  54     if seqLen < 1:
  55         print "Problems reading sequence from file"
  56
  57     print "writing to file %s" % chromOut
  58     outFile = open("%s%s" % (cisRoot, chromOut), "w")
  59     outFile.write(seq)
  60     outFile.close()
  61     ggGenome.addChromosomeEntry(chromID, chromOut, "file")
  62
  63
  64 def loadGeneEntries(db, gFile):
  65     """ FIXME - NEED TO DEAL WITH ALTERNATIVE SPLICING ENTRIES
  66     """
  67     geneEntries = []
  68     alreadySeen = []
  69     ggGenome = Genome("ggallus", dbFile=db)
  70     geneFile = open(gFile, "r")
  71     geneFile.readline()
  72     for line in geneFile:
  73         if "|" in line:
  74             continue
  75
  76         cols = line.split("\t")
  77         if cols[11].strip() != "GENE":
  78             continue
  79
  80         name = cols[10].split(":")
  81         gid = name[1]
  82         if gid == "" or gid in alreadySeen:
  83             continue
  84
  85         alreadySeen.append(gid)
  86         start = int(cols[2]) - 1
  87         stop = int(cols[3]) - 1
  88         sense = cols[4]
  89         chrom = cols[1].strip()
  90         if sense == "+":
  91             sense = "F"
  92         else:
  93             sense = "R"
  94
  95         geneID = ("ggallus", gid)
  96         gidVersion = 1
  97         geneEntries.append((geneID, chrom, start, stop, sense, "gene", gidVersion))
  98
  99     print "Adding %d gene entries" % len(geneEntries)
 100     ggGenome.addGeneEntryBatch(geneEntries)
 101
 102
 103 def loadGeneFeatures(db, gFile):
 104     """ Load gene features such as CDS, UTR, and PSEUDO from the gene file.
 105     """
 106     featureEntries = []
 107     ggGenome = Genome("ggallus", dbFile=db)
 108     featureFile = open(gFile, "r")
 109     featureFile.readline()
 110     for line in featureFile:
 111         if "|" in line:
 112             continue
 113
 114         cols = line.split("\t")
 115         if cols[11].strip() not in ["CDS", "UTR", "PSEUDO"]:
 116             continue
 117
 118         fType = cols[11]
 119         name = cols[10].split(":")
 120         gid = name[1]
 121         if gid == "":
 122             continue
 123
 124         start = int(cols[2]) - 1
 125         stop = int(cols[3]) - 1
 126         sense = cols[4]
 127         chrom = cols[1].strip()
 128         if sense == "+":
 129             sense = "F"
 130         else:
 131             sense = "R"
 132
 133         geneID = ("ggallus", gid)
 134         gidVersion = 1
 135         featureEntries.append((geneID, gidVersion, chrom, start, stop, sense, fType))
 136
 137     print "Adding %d feature entries" % len(featureEntries)
 138     ggGenome.addFeatureEntryBatch(featureEntries)
 139
 140
 141 def loadGeneOntology(db, goPath, goDefPath):
 142     ggGenome = Genome("ggallus", dbFile=db)
 143     goDefFile = open(goDefPath, "r")
 144     goFile = open(goPath, "r")
 145     idb = geneinfoDB()
 146     goDefs = {}
 147     goArray = []
 148     for goDefEntry in goDefFile:
 149         if goDefEntry[0] != "!":
 150             cols = goDefEntry.split("\t")
 151             goDefs[cols[0]] = (cols[1], cols[2].strip())
 152
 153     goEntries = goFile.readlines()
 154     prevGID = ""
 155     for entry in goEntries:
 156         try:
 157             fields = entry.split("\t")
 158             if fields[0] != "9031":
 159                 continue
 160
 161             locID = fields[1].strip()
 162             gID = ("ggallus", locID)
 163             if prevGID != gID:
 164                 prevGID = gID
 165                 gene_name = ""
 166                 synonyms = idb.geneIDSynonyms(gID)
 167                 if len(synonyms) >0:
 168                     gene_name = string.join(synonyms, ",")
 169
 170             goArray.append((gID, fields[2], "", gene_name, "", string.replace(goDefs[fields[2]][0], "'", "p"), goDefs[fields[2]][1], ""))
 171         except:
 172             print "locus ID %s could not be added" % locID
 173             pass
 174
 175     print "adding %d go entries" % len(goArray)
 176     ggGenome.addGoInfoBatch(goArray)
 177
 178
 179 def createDBFile(db):
 180     ggGenome = Genome("ggallus",  dbFile=db)
 181     ggGenome.createGeneDB(db)
 182
 183
 184 def createDBindices(db):
 185     ggGenome = Genome("ggallus", dbFile=db)
 186     ggGenome.createIndices()
 187
 188
 189 def buildChickenDB(db=geneDB):
 190     genePath = "%s/download/seq_gene.md" % cisRoot
 191     goDefPath = "%s/download/GO.terms_and_ids" % cisRoot # ftp://ftp.geneontology.org/pub/go/doc/GO.terms_and_ids
 192     goPath = "%s/download/gene2go" % cisRoot # ftp://ftp.ncbi.nih.gov/gene/DATA/gene2go.gz
 193     chromos = {"1": "%s/download/chr1.fa" % cisRoot,
 194                "2": "%s/download/chr2.fa" % cisRoot,
 195                "3": "%s/download/chr3.fa" % cisRoot,
 196                "4": "%s/download/chr4.fa" % cisRoot,
 197                "5": "%s/download/chr5.fa" % cisRoot,
 198                "6": "%s/download/chr6.fa" % cisRoot,
 199                "7": "%s/download/chr7.fa" % cisRoot,
 200                "8": "%s/download/chr8.fa" % cisRoot,
 201                "9": "%s/download/chr9.fa" % cisRoot,
 202                "10": "%s/download/chr10.fa" % cisRoot,
 203                "11": "%s/download/chr11.fa" % cisRoot,
 204                "12": "%s/download/chr12.fa" % cisRoot,
 205                "13": "%s/download/chr13.fa" % cisRoot,
 206                "14": "%s/download/chr14.fa" % cisRoot,
 207                "15": "%s/download/chr15.fa" % cisRoot,
 208                "16": "%s/download/chr16.fa" % cisRoot,
 209                "17": "%s/download/chr17.fa" % cisRoot,
 210                "18": "%s/download/chr18.fa" % cisRoot,
 211                "19": "%s/download/chr19.fa" % cisRoot,
 212                "20": "%s/download/chr20.fa" % cisRoot,
 213                "21": "%s/download/chr21.fa" % cisRoot,
 214                "22": "%s/download/chr22.fa" % cisRoot,
 215                "23": "%s/download/chr23.fa" % cisRoot,
 216                "24": "%s/download/chr24.fa" % cisRoot,
 217                "25": "%s/download/chr25.fa" % cisRoot,
 218                "26": "%s/download/chr26.fa" % cisRoot,
 219                "27": "%s/download/chr27.fa" % cisRoot,
 220                "28": "%s/download/chr28.fa" % cisRoot,
 221                "32": "%s/download/chr32.fa" % cisRoot,
 222                "W": "%s/download/chrW.fa" % cisRoot,
 223                "Z": "%s/download/chrZ.fa" % cisRoot,
 224                "M": "%s/download/chrM.fa" % cisRoot,
 225                "E22C19W28_E50C23": "%s/download/chrE22C19W28_E50C23.fa" % cisRoot,
 226                "E64": "%s/download/chrE64.fa" % cisRoot,
 227                "1_random": "%s/download/chr1_random.fa" % cisRoot,
 228                "2_random": "%s/download/chr2_random.fa" % cisRoot,
 229                "4_random": "%s/download/chr4_random.fa" % cisRoot,
 230                "5_random": "%s/download/chr5_random.fa" % cisRoot,
 231                "6_random": "%s/download/chr6_random.fa" % cisRoot,
 232                "7_random": "%s/download/chr7_random.fa" % cisRoot,
 233                "8_random": "%s/download/chr8_random.fa" % cisRoot,
 234                "10_random": "%s/download/chr10_random.fa" % cisRoot,
 235                "11_random": "%s/download/chr11_random.fa" % cisRoot,
 236                "12_random": "%s/download/chr12_random.fa" % cisRoot,
 237                "13_random": "%s/download/chr13_random.fa" % cisRoot,
 238                "16_random": "%s/download/chr16_random.fa" % cisRoot,
 239                "17_random": "%s/download/chr17_random.fa" % cisRoot,
 240                "18_random": "%s/download/chr18_random.fa" % cisRoot,
 241                "20_random": "%s/download/chr20_random.fa" % cisRoot,
 242                "22_random": "%s/download/chr22_random.fa" % cisRoot,
 243                "25_random": "%s/download/chr25_random.fa" % cisRoot,
 244                "28_random": "%s/download/chr28_random.fa" % cisRoot,
 245                "Un_random": "%s/download/chrUn_random.fa" % cisRoot,
 246                "W_random": "%s/download/chrW_random.fa" % cisRoot,
 247                "E64_random": "%s/download/chrE64_random.fa" % cisRoot,
 248                "Z_random": "%s/download/chrZ_random.fa" % cisRoot,
 249                "E22C19W28_E50C23_random": "%s/download/chrE22C19W28_E50C23_random.fa" % cisRoot
 250     }
 251
 252     print "Creating database %s" % db
 253     createDBFile(db)
 254
 255     print "Adding gene entries"
 256     loadGeneEntries(db, genePath)
 257
 258     #print "Adding gene annotations"
 259     #loadGeneAnnotations(db, annotPath)
 260
 261     print "Adding gene features"
 262     loadGeneFeatures(db, genePath)
 263
 264     for chromID in chromos.keys():
 265         print "Loading chromosome %s" % chromID
 266         loadChromosome(db, chromID, chromos[chromID], "/G_gallus/chromo%s.bin" % chromID)
 267
 268     print "Adding gene ontology"
 269     loadGeneOntology(db, goPath, goDefPath)
 270
 271     print "Creating Indices"
 272     createDBindices(db)
 273
 274     print "Finished creating database %s" % db