fix validate

valide.py → validate.py

@@ -68,6 +68,7 @@ def check_record_description(record_description):
         idSample = splitted[0]
         description = splitted[1]
         if '=' in idSample:
+            ## L'ID de la sequence ne doit pas comporter de caractere '=' car c'est attendu dans la description
             return False
         else:
             if "species_name=" in description:
@@ -75,13 +76,16 @@ def check_record_description(record_description):
                 descriptionRecord = DescriptionRecord(sampleid=idSample, species_name=speciesName)
                 return descriptionRecord
             else:
+                ## no species_name element in description
                 return False
     else:
+        ## the header doesn't have at least 2 colons separeted by ;
         return False
 
 def check_format_species_name(species_name):
     '''
     Check species name format
+    Excepted format "Genus species" or "Genus_species"
     '''
     patternSpeciesPerfect = "^[A-Z][a-z]+ [a-z]+$"
     if re.match(patternSpeciesPerfect, species_name):    
@@ -104,17 +108,20 @@ def check_dna_sequence(dna_sequence, code="ATGC"):
 
 
 def check_taxonomy_species_name(species_name, reference):
-    matching = { k:v for k,v in reference.items() if species_name in v }
+    matching = { k:v for k,v in reference.items() if species_name == v }
     if len(matching) == 1:
         rank = list(ncbi.get_rank(list(matching.keys())).values())[0]
         if rank == "species":
             return matching
         else:
-            return "the rank is not species but {0}".format(rank)
+            ## the rank in not "species"
+            return False
     elif len(matching) > 1:
-        return "more than 1 matching species: {0}".format(eval(str(matching)))
+        ## they are many matches so the species name is ambiguous
+        return False
     else:
-        return "species name not found in ncbi"
+        ## this species is unknown in ncbi
+        return False
 
 
 def full_taxonomy_sample(sampleid, taxonid):
@@ -132,6 +139,7 @@ def full_taxonomy_sample(sampleid, taxonid):
                 if idGenus and idFamily:
                     break
             if idGenus is False or idFamily is False:
+                ## Genus or Family is unknown in ncbi
                 return False
             else:
                 genusName = str(ncbi.translate_to_names([idGenus])[0])
@@ -147,8 +155,10 @@ def full_taxonomy_sample(sampleid, taxonid):
                                 rank = "species")
                 return sample
         else:
+            ## The taxon taxonomy is unknown or empty
             return False
     else:
+        ## This taxon have no taxonomy in ncbi
         return False
 
 
@@ -162,7 +172,7 @@ def full_taxonomy_sample(sampleid, taxonid):
 ##args
 localTaxdumpArchive = "TAXO/taxdump_2021.tar.gz"
 rawFastaFile="resources/test/raw.fasta"
-
+rawFastaFile="teleo_ok.fasta"
 
 ncbi = NCBITaxa()
 
@@ -177,7 +187,8 @@ vertDic = dict(zip_iterator)
 
 
 valideRecords = []
-faultyRecords = []
+faultyFormatRecords = []
+faultyTaxonRecords = []
 
 for record in SeqIO.parse(rawFastaFile, "fasta"):
     #print(record.id)
@@ -188,8 +199,9 @@ for record in SeqIO.parse(rawFastaFile, "fasta"):
         checkedFormat = check_format_species_name(checkedRecord.species_name)
         if checkedFormat is not False:
             if check_dna_sequence(record.seq):
-                checkedTaxon = check_taxonomy_species_name(checkedFormat, vertDic)
-                if isinstance(checkedTaxon, dict):
+                checkedRecord.species_name = checkedFormat
+                checkedTaxon = check_taxonomy_species_name(checkedRecord.species_name, vertDic)
+                if checkedTaxon is not False:
                     #print(checkedTaxon)
                     sample = full_taxonomy_sample(checkedRecord.sampleid , list(checkedTaxon.keys())[0])
                     if sample is not False:
@@ -197,33 +209,33 @@ for record in SeqIO.parse(rawFastaFile, "fasta"):
                         valideRecords.append(thisValideRecord)
                     else:
                         ## faulty rank taxonomy
-                        faultydescription = str(record.description)+' | faulty rank taxonomy: family or genera not found'
+                        faultydescription = str(record.description)+'; faulty rank taxonomy: family or genera not found'
                         thisFaultyRecord = SeqRecord(id = record.id, description = faultydescription, seq=record.seq)
-                        faultyRecords.append(thisFaultyRecord)
+                        faultyTaxonRecords.append(thisFaultyRecord)
                 else:
                     ## faulty taxonomy
-                    faultydescription = str(record.description)+' | faulty taxonomy '+str(checkedTaxon)
+                    faultydescription = str(record.description)+'; faulty taxonomy: species name '+ str(checkedRecord.species_name)+ ' not found in NCBI'
                     thisFaultyRecord = SeqRecord(id = record.id, description = faultydescription, seq=record.seq)
-                    faultyRecords.append(thisFaultyRecord)
+                    faultyTaxonRecords.append(thisFaultyRecord)
             else:
-                faultydescription = str(record.description)+' | faulty DNA sequence'
+                faultydescription = str(record.description)+' ; faulty DNA sequence'
                 thisFaultyRecord = SeqRecord(id = record.id, description = faultydescription, seq=record.seq)
-                faultyRecords.append(thisFaultyRecord)
+                faultyFormatRecords.append(thisFaultyRecord)
         else:
             ## faulty species name format
-            faultydescription = str(record.description)+' | faulty species name format ' + str(checkedRecord.species_name)
+            faultydescription = str(record.description)+' ; faulty species name format ' + str(checkedRecord.species_name)
             thisFaultyRecord = SeqRecord(id = record.id, description = faultydescription, seq=record.seq)
-            faultyRecords.append(thisFaultyRecord)
+            faultyFormatRecords.append(thisFaultyRecord)
     else:
         ## faulty record description format
         print("ERROR FORMAT FASTA {0}: {1} {2}.\n Excepted format is\n\n>sampleid ; species_names=Genus Species\nACTAG\n".format(rawFastaFile,record.id, record.description))
 
-
-
+## write faulty format records fasta
+SeqIO.write(faultyFormatRecords, "faulty_format.fasta", "fasta")
+## write faulty format records fasta
+SeqIO.write(faultyTaxonRecords, "faulty_taxon.fasta", "fasta")
 ## write valide records fasta
 SeqIO.write(valideRecords, "valide.fasta", "fasta")
-## write faulty records fasta
-SeqIO.write(faultyRecords, "faulty.fasta", "fasta")