Merge pull request #2376 from merenlab/iss2375

A global argument to keep all functional annotations per gene (fixes #2375)

anvio/__init__.py

@@ -74,6 +74,9 @@
 USER_KNOWS_IT_IS_NOT_A_GOOD_IDEA = '--I-know-this-is-not-a-good-idea' in sys.argv
 DOCS_PATH = os.path.join(os.path.dirname(__file__), 'docs')
 TMP_DIR = None
+# global args that we can set internally as needed
+RETURN_ALL_FUNCTIONS_FROM_SOURCE_FOR_EACH_GENE = False  # set to True if you want all functional annotations from a given annotation source
+                                                        # instead of the best hit per gene
 
 # if the user wants to use a non-default tmp directory, we set it here
 if '--tmp-dir' in sys.argv:

anvio/dbops.py

@@ -602,6 +602,9 @@ def init_functions(self, requested_sources=[], dont_panic=False):
         If self.split_names_of_interest has a value, the dictionary only includes gene calls from those splits.
 
         Afterwards, it sets self.gene_function_calls_initiated to True.
+
+        Note: the global argument RETURN_ALL_FUNCTIONS_FROM_SOURCE_FOR_EACH_GENE affects the behavior of this function. If False, we get 
+        the best hit per gene (lowest e-value) for a given annotation source. If True, we get all hits.
         """
         if not self.contigs_db_path:
             return
@@ -643,13 +646,21 @@ def init_functions(self, requested_sources=[], dont_panic=False):
             if gene_callers_id not in self.gene_function_calls_dict:
                 self.gene_function_calls_dict[gene_callers_id] = dict([(s, None) for s in self.gene_function_call_sources])
 
-            if self.gene_function_calls_dict[gene_callers_id][source] and e_value:
-                if self.gene_function_calls_dict[gene_callers_id][source][2] < e_value:
-                    # 'what we have:', self.gene_function_calls_dict[gene_callers_id][source]
-                    # 'rejected    :', ('%s :: %s' % (function if function else 'unknown', accession), e_value)
-                    continue
-
             entry = (accession, '%s' % (function if function else 'unknown'), e_value)
+
+            if self.gene_function_calls_dict[gene_callers_id][source]:
+                if anvio.RETURN_ALL_FUNCTIONS_FROM_SOURCE_FOR_EACH_GENE:
+                    previous_entry_acc, previous_entry_func, previous_entry_evalue = self.gene_function_calls_dict[gene_callers_id][source]
+                    combined_acc = f"{previous_entry_acc}!!!{accession}"
+                    combined_func = f"{previous_entry_func}!!!{entry[1]}"
+                    combined_evalue = f"{previous_entry_evalue}!!!{e_value}"
+                    entry = (combined_acc, combined_func, combined_evalue)
+                else:
+                    if e_value and self.gene_function_calls_dict[gene_callers_id][source][2] < e_value:
+                        # 'what we have:', self.gene_function_calls_dict[gene_callers_id][source]
+                        # 'rejected    :', ('%s :: %s' % (function if function else 'unknown', accession), e_value)
+                        continue
+
             self.gene_function_calls_dict[gene_callers_id][source] = entry
 
         contigs_db.disconnect()

bin/anvi-summarize

@@ -62,6 +62,9 @@ def main(args):
         if not args.contigs_db:
             raise ConfigError("You must provide a contigs database when you summarize anvi'o profiles. True story.")
 
+        # set this global arg so that all functional annotations are reflected in the summary output
+        anvio.RETURN_ALL_FUNCTIONS_FROM_SOURCE_FOR_EACH_GENE = True
+
         summary = summarizer.ProfileSummarizer(args)
     else:
         raise ConfigError("Well. '%s' is neither an anvi'o pan database, nor an anvi'o profile database. There is nothing this "