Add new algorithm in EcoTag to reduce impact of database errors

src/ecoTag.py

@@ -214,7 +214,7 @@ if __name__=='__main__':
 
         if best[0] is not None:
             taxlist = set(taxonlink[p[0].id] for p in match)
-            lca = taxonomy.lastCommonTaxon(*tuple(taxlist))
+            lca = taxonomy.betterCommonTaxon(0.2,*tuple(taxlist))
             scname = taxonomy.getScientificName(lca)
             rank = taxonomy.getRank(lca)
       

src/obitools/ecopcr/taxonomy.py

@@ -1,11 +1,12 @@
 import struct
 import sys
 
-from itertools import count,imap
+from itertools import count,imap,combinations
 
 from obitools.ecopcr import EcoPCRDBFile
 from obitools.utils import universalOpen
 from obitools.utils import ColumnFile
+import math
 
 class Taxonomy(object):
     def __init__(self):
@@ -116,11 +117,21 @@ class Taxonomy(object):
             
         return ancetre
     
-    def betterCommonTaxon(self,error=1,*taxids):
+    def depth(self,taxid):
+        return len([x for x in self.parentalTreeIterator(taxid)])
+    
+    def betterCommonTaxon(self,error=0.2,*taxids): 
+        
+        error = math.floor(len(taxids) * error)
+        
+        if error >= 1:
+            possible = [self.lastCommonTaxon(x) for x in combinations(taxids,len(taxids)-error)]
+            possible.sort(cmp=lambda t1,t2: cmp(self.depth(t2),self.depth(t1)))
+            lca=possible[0]
+        else:
             lca = self.lastCommonTaxon(*taxids)
-        idx = self._index[lca]
+           
-        sublca = [t[0] for t in self._taxonomy if t[2]==idx]
+        return lca
-        return sublca
 
     
     def getPreferedName(self,taxid):