Merge pull request #486 from TranslatorSRI/add-geneprotein-conflated-synonyms

This PR adds a new output: `synonyms/GeneProteinConflated.txt`, based on the GeneProtein conflation and the Gene and Protein outputs. It should be pretty similar to the first attempt (PR #185). Also fixes a bug in synonym conflating with taxa that we'd never run into because we'd only conflated DrugChemicals before.

Closes NCATSTranslator/NameResolution#191.

Author: gaurav

config.yaml

@@ -289,6 +289,9 @@ chemical_outputs:
 drugchemicalconflated_synonym_outputs:
   - DrugChemicalConflated.txt
 
+geneproteinconflated_synonym_outputs:
+  - GeneProteinConflated.txt
+
 taxon_labels:
   - NCBITaxon
   - MESH

src/snakefiles/geneprotein.snakefile

@@ -1,5 +1,6 @@
 import src.createcompendia.geneprotein as geneprotein
-import src.assess_compendia as assessments
+from src.synonyms import synonymconflation
+from util import gzip_files
 
 ### Gene / Protein
 
@@ -22,9 +23,26 @@ rule geneprotein_conflation:
     run:
         geneprotein.build_conflation(input.geneprotein_concord,input.gene_compendium,input.protein_compendium,output.outfile)
 
+rule geneprotein_conflated_synonyms:
+    input:
+        geneprotein_conflations=[config['output_directory']+'/conflation/GeneProtein.txt'],
+        gene_compendia=expand("{od}/compendia/{ap}", od = config['output_directory'], ap = config['gene_outputs']),
+        protein_compendia=expand("{od}/compendia/{ap}", od = config['output_directory'], ap = config['protein_outputs']),
+        gene_synonyms_gz=expand("{od}/synonyms/{ap}.gz", od = config['output_directory'], ap = config['gene_outputs']),
+        protein_synonyms_gz=expand("{od}/synonyms/{ap}.gz", od = config['output_directory'], ap = config['protein_outputs'])
+    output:
+        geneprotein_conflated_synonyms_gz=config['output_directory']+'/synonyms/GeneProteinConflated.txt.gz'
+    run:
+        synonymconflation.conflate_synonyms(
+            input.gene_synonyms_gz + input.protein_synonyms_gz,
+            input.gene_compendia + input.protein_compendia,
+            input.geneprotein_conflations,
+            output.geneprotein_conflated_synonyms_gz)
+
 rule geneprotein:
     input:
-        config['output_directory']+'/conflation/GeneProtein.txt'
+        config['output_directory']+'/conflation/GeneProtein.txt',
+        config['output_directory']+'/synonyms/GeneProteinConflated.txt.gz'
     output:
         x=config['output_directory']+'/reports/geneprotein_done'
     shell:

src/snakefiles/util.py

@@ -62,6 +62,7 @@ def get_all_synonyms(config):
         config['cell_line_outputs'] +
         config['genefamily_outputs'] +
         config['drugchemicalconflated_synonym_outputs'] +
+        config['geneproteinconflated_synonym_outputs'] +
         config['umls_outputs'] +
         config['macromolecularcomplex_outputs'] +
         # Publication.txt is empty, but it's still created, so it needs to be here.
@@ -87,6 +88,7 @@ def get_all_synonyms_except_drugchemicalconflated(config):
             config['cell_line_outputs'] +
             config['genefamily_outputs'] +
             # config['drugchemicalconflated_synonym_outputs'] +
+            config['geneproteinconflated_synonym_outputs'] +
             config['umls_outputs'] +
             config['macromolecularcomplex_outputs']
     )
@@ -110,6 +112,7 @@ def get_all_synonyms_with_drugchemicalconflated(config):
             config['cell_line_outputs'] +
             config['genefamily_outputs'] +
             config['drugchemicalconflated_synonym_outputs'] +
+            config['geneproteinconflated_synonym_outputs'] +
             config['umls_outputs'] +
             config['macromolecularcomplex_outputs']
     )

src/synonyms/synonymconflation.py

@@ -213,7 +213,7 @@ def conflate_synonyms(synonym_files_gz, compendia_files, conflation_file, output
                         if 'taxa' in synonym:
                             if 'taxa' not in final_conflation:
                                 final_conflation['taxa'] = set()
-                            final_conflation.update(synonym['taxa'])
+                            final_conflation['taxa'].update(synonym['taxa'])
 
             # Convert the taxa into a list.
             final_conflation['taxa'] = sorted(final_conflation['taxa'])