remove non-standard gene symbols in comliant mode

oschwengers · Nov 13, 2024 · 30b3928 · 30b3928
1 parent 7aa8ebc
commit 30b3928
Show file tree

Hide file tree

Showing 2 changed files with 14 additions and 2 deletions.
diff --git a/bakta/io/gff.py b/bakta/io/gff.py
@@ -8,6 +8,7 @@
 import bakta.constants as bc
 import bakta.io.fasta as fasta
 import bakta.io.insdc as insdc
+import bakta.features.annotation as ba
 import bakta.so as so
 
 
@@ -165,6 +166,10 @@ def write_features(data: dict, features_by_sequence: Dict[str, dict], gff3_path:
                             'locus_tag': feat['locus'],
                             'gene': feat['gene']
                         }
+                        if(ba.RE_GENE_SYMBOL.fullmatch(feat['gene'])):  # discard non-standard ncRNA gene symbols
+                            gene_annotations['gene'] = feat['gene']
+                        else:
+                            annotations.pop('gene')
                         if('truncated' in feat):
                             gene_annotations[bc.INSDC_FEATURE_PSEUDO] = True
                         gene_annotations = encode_annotations(gene_annotations)

diff --git a/bakta/io/insdc.py b/bakta/io/insdc.py
@@ -12,6 +12,7 @@
 import bakta
 import bakta.config as cfg
 import bakta.constants as bc
+import bakta.features.annotation as ba
 import bakta.so as so
 
 
@@ -258,8 +259,14 @@ def build_biopython_sequence_list(data: dict, features: Sequence[dict]):
                     'locus_tag': feature['locus']
                 }
                 if(feature.get('gene', None)):
-                    qualifiers['gene'] = feature['gene']
-                    gene_qualifier['gene'] = feature['gene']
+                    if(cfg.compliant):
+                        if(ba.RE_GENE_SYMBOL.fullmatch(feature['gene'])):  # discard non-standard gene symbols
+                            gene_qualifier['gene'] = feature['gene']
+                        else:
+                            qualifiers.pop('gene')
+                    else:
+                        qualifiers['gene'] = feature['gene']
+                        gene_qualifier['gene'] = feature['gene']
                 if(bc.PSEUDOGENE in feature):
                     if(feature['type'] == bc.FEATURE_CDS):
                         gene_qualifier[bc.INSDC_FEATURE_PSEUDOGENE] = bc.INSDC_FEATURE_PSEUDOGENE_TYPE_UNPROCESSED if feature[bc.PSEUDOGENE]['paralog'] else bc.INSDC_FEATURE_PSEUDOGENE_TYPE_UNITARY