cistematic/genomes/cfamiliaris.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 Canis familiaris
  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/C_familiaris/cfamiliaris.genedb" % cisRoot
  41
  42
  43 def loadChromosome(db, chromID, chromPath, chromOut):
  44     seqArray = []
  45     cfGenome = Genome("cfamiliaris", 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     cfGenome.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     cfGenome = Genome("cfamiliaris", 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 = ("cfamiliaris", gid)
  92         gidVersion = 1
  93         geneEntries.append((geneID, chrom, start, stop, sense, "gene", gidVersion))
  94
  95     print "Adding %d gene entries" % len(geneEntries)
  96     cfGenome.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     cfGenome = Genome("cfamiliaris", 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 = ("cfamiliaris", gid)
 127         gidVersion = 1
 128         featureEntries.append((geneID, gidVersion, chrom, start, stop, sense, fType))
 129
 130     print "Adding %d feature entries" % len(featureEntries)
 131     cfGenome.addFeatureEntryBatch(featureEntries)
 132
 133
 134 def loadGeneAnnotations(db, annotPath):
 135     geneAnnotations = []
 136     annotFile = open(annotPath, "r")
 137     cfGenome = Genome("cfamiliaris", dbFile=db)
 138     for line in annotFile:
 139         try:
 140             cols = line.split("\t")
 141             locID = cols[0]
 142             geneDesc = cols[6]
 143             if len(locID) > 0:
 144                 geneAnnotations.append((("cfamiliaris", locID), string.replace(geneDesc.strip(), "'", "p")))
 145         except:
 146             pass
 147
 148     print "Adding %d annotations" % len(geneAnnotations)
 149     cfGenome.addAnnotationBatch(geneAnnotations)
 150
 151
 152 def loadGeneOntology(db, goPath, goDefPath):
 153     cfGenome = Genome("cfamiliaris", dbFile=db)
 154     goDefFile = open(goDefPath, "r")
 155     goFile = open(goPath, "r")
 156     idb = geneinfoDB()
 157     goDefs = {}
 158     goArray = []
 159     for goDefEntry in goDefFile:
 160         if goDefEntry[0] != "!":
 161             cols = goDefEntry.split("\t")
 162             goDefs[cols[0]] = (cols[1], cols[2].strip())
 163
 164     goEntries = goFile.readlines()
 165     prevGID = ''
 166     for entry in goEntries:
 167         try:
 168             fields = entry.split("\t")
 169             if fields[0] != "9615":
 170                 continue
 171
 172             locID = fields[1].strip()
 173             gID = ("cfamiliaris", locID)
 174             if prevGID != gID:
 175                 prevGID = gID
 176                 gene_name = ""
 177                 synonyms = idb.geneIDSynonyms(gID)
 178                 if len(synonyms) >0:
 179                     for entry in synonyms:
 180                         gene_name += ","
 181                         gene_name += entry
 182                 else:
 183                     gene_name = " "
 184
 185             goArray.append((gID, fields[2], "", gene_name[1:], "", string.replace(goDefs[fields[2]][0], "'", "p"), goDefs[fields[2]][1], ""))
 186         except:
 187             print "locus ID %s could not be added" % locID
 188             pass
 189
 190     print "adding %d go entries" % len(goArray)
 191     cfGenome.addGoInfoBatch(goArray)
 192
 193
 194 def createDBFile(db):
 195     cfGenome = Genome("cfamiliaris",  dbFile=db)
 196     cfGenome.createGeneDB(db)
 197
 198
 199 def createDBindices(db):
 200     cfGenome = Genome("cfamiliaris", dbFile=db)
 201     cfGenome.createIndices()
 202
 203
 204 def buildDogDB(db=geneDB):
 205     genePath = "%s/download/seq_gene.md" % cisRoot
 206     chromos = {"1": "%s/download/chr1.fa" % cisRoot,
 207                "2": "%s/download/chr2.fa" % cisRoot,
 208                "3": "%s/download/chr3.fa" % cisRoot,
 209                "4": "%s/download/chr4.fa" % cisRoot,
 210                "5": "%s/download/chr5.fa" % cisRoot,
 211                "6": "%s/download/chr6.fa" % cisRoot,
 212                "7": "%s/download/chr7.fa" % cisRoot,
 213                "8": "%s/download/chr8.fa" % cisRoot,
 214                "9": "%s/download/chr9.fa" % cisRoot,
 215                "10": "%s/download/chr10.fa" % cisRoot,
 216                "11": "%s/download/chr11.fa" % cisRoot,
 217                "12": "%s/download/chr12.fa" % cisRoot,
 218                "13": "%s/download/chr13.fa" % cisRoot,
 219                "14": "%s/download/chr14.fa" % cisRoot,
 220                "15": "%s/download/chr15.fa" % cisRoot,
 221                "16": "%s/download/chr16.fa" % cisRoot,
 222                "17": "%s/download/chr17.fa" % cisRoot,
 223                "18": "%s/download/chr18.fa" % cisRoot,
 224                "19": "%s/download/chr19.fa" % cisRoot,
 225                "20": "%s/download/chr20.fa" % cisRoot,
 226                "21": "%s/download/chr21.fa" % cisRoot,
 227                "22": "%s/download/chr22.fa" % cisRoot,
 228                '23': "%s/download/chr23.fa" % cisRoot,
 229                "24": "%s/download/chr24.fa" % cisRoot,
 230                "25": "%s/download/chr25.fa" % cisRoot,
 231                "26": "%s/download/chr26.fa" % cisRoot,
 232                "27": "%s/download/chr27.fa" % cisRoot,
 233                "28": "%s/download/chr28.fa" % cisRoot,
 234                "29": "%s/download/chr29.fa" % cisRoot,
 235                "30": "%s/download/chr30.fa" % cisRoot,
 236                "31": "%s/download/chr31.fa" % cisRoot,
 237                "32": "%s/download/chr32.fa" % cisRoot,
 238                "33": "%s/download/chr33.fa" % cisRoot,
 239                "34": "%s/download/chr34.fa" % cisRoot,
 240                "35": "%s/download/chr35.fa" % cisRoot,
 241                "36": "%s/download/chr36.fa" % cisRoot,
 242                "37": "%s/download/chr37.fa" % cisRoot,
 243                "38": "%s/download/chr38.fa" % cisRoot,
 244                "X": "%s/download/chrX.fa" % cisRoot,
 245                "Un": "%s/download/chrUn.fa" % cisRoot
 246     }
 247
 248     print "Creating database %s" % db
 249     createDBFile(db)
 250
 251     print "Adding gene entries"
 252     loadGeneEntries(db, genePath)
 253
 254     print "Adding gene features"
 255     loadGeneFeatures(db, genePath)
 256
 257     for chromID in chromos.keys():
 258         print "Loading chromosome %s" % chromID
 259         loadChromosome(db, chromID, chromos[chromID], "/C_familiaris/chromo%s.bin" % chromID)
 260
 261     print "Creating Indices"
 262     createDBindices(db)
 263
 264     print "Finished creating database %s" % db