Do not crash on non-numeric locus tag end, fixes #81

git-svn-id: https://humgenprojects.lumc.nl/svn/mutalyzer/trunk@452 eb6bd6ab-9ccd-42b9-aceb-e2899b4a52f1

mutalyzer/parsers/genbank.py

@@ -4,6 +4,7 @@ mutalyzer GenRecord. Record populated with data from a GenBank file.
 """
 
 
+import re
 import bz2
 from itertools import izip_longest
 
@@ -15,6 +16,10 @@ from mutalyzer import Db
 from mutalyzer.GenRecord import PList, Locus, Gene, Record, GenRecord
 
 
+# Regular expression used to find version number in locus tag
+LOCUS_TAG_VERSION = re.compile('\d{1,3}$')
+
+
 class tempGene():
     """
     Container class for a given gene name.
@@ -564,9 +569,15 @@ class GBparser():
                             # Normally, locus_tag ends with three digits, but
                             # for some (e.g. mobA on NC_011228, a plasmid) it
                             # ends with two digits prepended with an
-                            # underscore. We really want a number, so 'fix'
-                            # this by substituting a zero.
-                            myTranscript = Locus(i.locus_tag[-3:].replace('_', '0'))
+                            # underscore. Or prepended with a letter. We
+                            # really want a number, so 'fix' this by only
+                            # looking for a numeric part.
+                            try:
+                                version = LOCUS_TAG_VERSION.findall(
+                                    i.locus_tag)[0].zfill(3)
+                            except IndexError:
+                                version = '000'
+                            myTranscript = Locus(version)
                         else :
                             myTranscript = Locus(myRealGene.newLocusTag())
                         myTranscript.mRNA = PList()
@@ -603,9 +614,15 @@ class GBparser():
                             # Normally, locus_tag ends with three digits, but
                             # for some (e.g. mobA on NC_011228, a plasmid) it
                             # ends with two digits prepended with an
-                            # underscore. We really want a number, so 'fix'
-                            # this by substituting a zero.
-                            myTranscript = Locus(i.locus_tag[-3:].replace('_', '0'))
+                            # underscore. Or prepended with a letter. We
+                            # really want a number, so 'fix' this by only
+                            # looking for a numeric part.
+                            try:
+                                version = LOCUS_TAG_VERSION.findall(
+                                    i.locus_tag)[0].zfill(3)
+                            except IndexError:
+                                version = '000'
+                            myTranscript = Locus(version)
                         else :
                             myTranscript = Locus(myRealGene.newLocusTag())
                         myTranscript.CDS = PList()

tests/test_variantchecker.py

@@ -453,3 +453,10 @@ class TestVariantchecker():
         assert_equal(error_count, 0)
         assert_equal(self.output.getIndexedOutput('genomicDescription', 0),
                      'LRG_1:g.6855G>T')
+
+    def test_non_numeric_locus_tag_ending(self):
+        """
+        Locus tag in NC_002128 does not end in an underscore and three digits
+        but we should not crash on it.
+        """
+        check_variant('NC_002128(tagA):c.3del', self.output)