Update format_ncbi_taxonomy.py

add quotes around name for generated enum
datacommonsorg · Jul 4, 2024 · 9a9fb70 · 9a9fb70
1 parent 26f8251
commit 9a9fb70
Showing 1 changed file with 12 additions and 6 deletions.
diff --git a/scripts/biomedical/NCBI_Taxonomy/scripts/format_ncbi_taxonomy.py b/scripts/biomedical/NCBI_Taxonomy/scripts/format_ncbi_taxonomy.py
@@ -14,6 +14,8 @@
 """
 Author: Pradeep Kumar Krishnaswamy
 Date: 18-Apr-2024
+Edited By: Samantha Piekos
+Last Edited: 03-Jul-2024
 Name: format_ncbi_taxonomy
 Description: cleaning the NCBI Taxonomy data.
 @source data: Download Taxdump.tar.z from NCBI Taxonomy FTP Download page
@@ -33,13 +35,14 @@
 from absl import flags
 from absl import logging
 
-SOURCE_FILE_PATH = None
-OUTPUT_FILE_PATH = None
+SOURCE_FILE_PATH = 'input/'
+OUTPUT_FILE_PATH = 'output/'
 OUTPUT_MCF_FILE = 'ncbi_taxonomy_schema_enum.mcf'
 OUTPUT_TAXID_DCID_MAPPING_FILE = 'tax_id_dcid_mapping.txt'
 OUTPUT_NCBI_TAXONOMY_CSV = 'ncbi_taxonomy.csv'
 
-FIRST_MCF_ENTRY = """Node:dcid:BiologicalTaxonomicDivisionEnum
+FIRST_MCF_ENTRY = """# This schema file is generated by format_ncbi_taxonomy.py
+Node:dcid:BiologicalTaxonomicDivisionEnum
 name: "BiologicalTaxonomicDivisionEnum"
 typeOf: schema:Class
 subClassOf: schema:Enumeration
@@ -50,7 +53,7 @@
 DIVISION_COL = [
     'division_code', 'division_acronym', 'division_name', 'comments'
 ]
-DIVISION_MCF = "Node:dcid:BiologicalTaxonomicDivision{PascalCaseDivisionName}\nname: {division_name}\ntypeOf: dcs:BiologicalTaxonomicDivisionEnum\nacronym: {division_acronym}\ndescription: {comments}\n\n"  # pylint: disable=line-too-long
+DIVISION_MCF = 'Node:dcid:BiologicalTaxonomicDivision{PascalCaseDivisionName}\nname: "{division_name}"\ntypeOf: dcs:BiologicalTaxonomicDivisionEnum\nabbreviation: "{division_acronym}"\ndescription: "{comments}"\n\n'  # pylint: disable=line-too-long
 
 DIVISION_DICT = {}
 NCBI_TAXONOMY_CSV_HEADER = [
@@ -158,7 +161,7 @@
 subClassOf: schema:Enumeration
 description: "A host is a larger organism that harbors a smaller organism. The relationship between the two organisms can be parasitic, mutualistic, or commensalist. This encodes the type of larger organism that is serving as a host."
 \n"""
-HOST_MCF = "Node:dcid:BiologicalHost{biological_case_item}\nname: {item}\ntypeOf: dcs:BiologicalHostEnum\n\n"  # pylint: disable=line-too-long
+HOST_MCF = 'Node:dcid:BiologicalHost{biological_case_item}\nname: "{item}"\ntypeOf: dcs:BiologicalHostEnum\n\n'  # pylint: disable=line-too-long
 HOST_DICT = {}
 
 NODES_DMP = 'nodes.dmp'
@@ -174,7 +177,7 @@
 descriptionUrl: "https://en.wikipedia.org/wiki/Taxonomic_rank"
 \n
 """
-NODES_MCF = "Node:dcid:BiologicalTaxonomicRank{rank_case}\nname: {rank}\ntypeOf: dcs:BiologicalTaxonomicRankEnum\n\n"  # pylint: disable=line-too-long
+NODES_MCF = 'Node:dcid:BiologicalTaxonomicRank{rank_case}\nname: "{rank}"\ntypeOf: dcs:BiologicalTaxonomicRankEnum\n\n'  # pylint: disable=line-too-long
 NODES_DICT = {}
 
 CATEGORIES_DMP = 'categories.dmp'
@@ -511,6 +514,8 @@ def update_nodes_enum(self, df: pd.DataFrame) -> None:
                 current_node = FINAL_NCBI_TAXONOMY[row['tax_id']]
                 current_node['parentDcid'] = TAX_ID_DCID_MAPPING[
                     row["parent_tax_id"]]
+                rank_case = string.capwords(row["rank"]).replace(" ", "")
+                current_node['taxonRank'] = "dcs:BiologicalTaxonomicRank" + rank_case
                 current_node['division'] = DIVISION_DICT[row["division_code"]]
                 current_node['hasInheritedDivision'] = True if row[
                     'inherited_division'] == 1 else False
@@ -654,3 +659,4 @@ def main(_):
 
 if __name__ == "__main__":
     app.run(main)
+