From 2baa88b766ec540bd34b96599014dd16e393af39 Mon Sep 17 00:00:00 2001
From: AndreaGuarracino
Date: Tue, 25 Aug 2020 00:02:32 +0200
Subject: the YAML/FASTA pair is not created for samples where at least one
 mandatory field is missing

---
 .../from_genbank_to_fasta_and_yaml.py              | 34 ++++++++++++++++++----
 1 file changed, 29 insertions(+), 5 deletions(-)

(limited to 'scripts/download_genbank_data')

diff --git a/scripts/download_genbank_data/from_genbank_to_fasta_and_yaml.py b/scripts/download_genbank_data/from_genbank_to_fasta_and_yaml.py
index f314a1d..eefdddb 100755
--- a/scripts/download_genbank_data/from_genbank_to_fasta_and_yaml.py
+++ b/scripts/download_genbank_data/from_genbank_to_fasta_and_yaml.py
@@ -139,7 +139,7 @@ min_len_to_count = 27500
 num_seq_with_len_ge_X_bp = 0
 
 missing_value_list = []
-not_created_accession_list = []
+not_created_accession_dict = {}
 accession_with_errors_list = []
 
 for path_metadata_xxx_xml in [os.path.join(dir_metadata, name_metadata_xxx_xml) for name_metadata_xxx_xml in os.listdir(dir_metadata) if name_metadata_xxx_xml.endswith('.xml')]:
@@ -374,10 +374,34 @@ for path_metadata_xxx_xml in [os.path.join(dir_metadata, name_metadata_xxx_xml)
                     elif GBQualifier_name_text == 'db_xref':
                         info_for_yaml_dict['virus']['virus_species'] = "http://purl.obolibrary.org/obo/NCBITaxon_"+GBQualifier_value_text.split('taxon:')[1]
 
-
+            # Check if mandatory fields are missing
             if 'sample_sequencing_technology' not in info_for_yaml_dict['technology']:
                 #print(accession_version, ' - technology not found')
-                not_created_accession_list.append([accession_version, 'technology not found'])
+                if accession_version not in not_created_accession_dict:
+                    not_created_accession_dict[accession_version] = []
+                not_created_accession_dict[accession_version].append('sample_sequencing_technology not found')
+
+            if 'collection_location' not in info_for_yaml_dict['sample']:
+                if accession_version not in not_created_accession_dict:
+                    not_created_accession_dict[accession_version] = []
+                not_created_accession_dict[accession_version].append('collection_location not found')
+
+            if 'collection_date' not in info_for_yaml_dict['sample']:
+                if accession_version not in not_created_accession_dict:
+                    not_created_accession_dict[accession_version] = []
+                not_created_accession_dict[accession_version].append('collection_date not found')
+
+            if 'authors' not in info_for_yaml_dict['submitter']:
+                if accession_version not in not_created_accession_dict:
+                    not_created_accession_dict[accession_version] = []
+                not_created_accession_dict[accession_version].append('authors not found')
+
+            if 'host_species' not in info_for_yaml_dict['host']:
+                if accession_version not in not_created_accession_dict:
+                    not_created_accession_dict[accession_version] = []
+                not_created_accession_dict[accession_version].append('host_species not found')
+
+            if accession_version in not_created_accession_dict:
                 continue
 
             with open(os.path.join(dir_fasta_and_yaml, '{}.fasta'.format(accession_version)), 'w') as fw:
@@ -406,10 +430,10 @@ if len(accession_with_errors_list) > 0:
     with open(path_accession_with_errors_tsv, 'w') as fw:
         fw.write('\n'.join(accession_with_errors_list))
 
-if len(not_created_accession_list) > 0:
+if len(not_created_accession_dict) > 0:
     path_not_created_accession_tsv = 'not_created_accession.genbank.tsv'
     print('Written not created accession in {}'.format(path_not_created_accession_tsv))
     with open(path_not_created_accession_tsv, 'w') as fw:
-        fw.write('\n'.join(['\t'.join(x) for x in not_created_accession_list]))
+        fw.write('\n'.join(['\t'.join([accession_version, ','.join(missing_info_list)]) for accession_version, missing_info_list in not_created_accession_dict.items()]))
 
 print('Num. new sequences with length >= {} bp: {}'.format(min_len_to_count, num_seq_with_len_ge_X_bp))
-- 
cgit v1.2.3