aboutsummaryrefslogtreecommitdiff
path: root/scripts/download_genbank_data
diff options
context:
space:
mode:
authorAndreaGuarracino2020-08-27 00:18:24 +0200
committerAndreaGuarracino2020-08-27 00:18:24 +0200
commit4299c750728bbad4bdbf0311ff2a4b9c65d9883c (patch)
tree07515a9ad6d07eca22853806f879fb39d2375c09 /scripts/download_genbank_data
parentb69b8560132786e16ea6a997d65fd8c43381cc03 (diff)
downloadbh20-seq-resource-4299c750728bbad4bdbf0311ff2a4b9c65d9883c.tar.gz
bh20-seq-resource-4299c750728bbad4bdbf0311ff2a4b9c65d9883c.tar.lz
bh20-seq-resource-4299c750728bbad4bdbf0311ff2a4b9c65d9883c.zip
updated dependency from clustalw to minimap2; the genbank script no longer creates YAML/FASTA pairs for too short sequences
Diffstat (limited to 'scripts/download_genbank_data')
-rwxr-xr-xscripts/download_genbank_data/from_genbank_to_fasta_and_yaml.py19
1 files changed, 9 insertions, 10 deletions
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 272b5ba..8ef76e1 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
@@ -145,7 +145,7 @@ for path_dict_xxx_csv in [os.path.join(dir_dict_ontology_standardization, name_x
if not os.path.exists(dir_fasta_and_yaml):
os.makedirs(dir_fasta_and_yaml)
-min_len_to_count = 27500
+min_len_to_count = 15000
num_seq_with_len_ge_X_bp = 0
missing_value_list = []
@@ -411,18 +411,17 @@ for path_metadata_xxx_xml in [os.path.join(dir_metadata, name_metadata_xxx_xml)
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:
- fw.write('>{}\n{}'.format(accession_version, GBSeq_sequence.text.upper()))
+ if len(GBSeq_sequence.text) < min_len_to_count:
+ not_created_accession_dict[accession_version].append('sequence shorter than {} bp'.format(min_len_to_count))
- with open(os.path.join(dir_fasta_and_yaml, '{}.yaml'.format(accession_version)), 'w') as fw:
- json.dump(info_for_yaml_dict, fw, indent=2)
+ if accession_version not in not_created_accession_dict:
+ num_seq_with_len_ge_X_bp += 1
+ with open(os.path.join(dir_fasta_and_yaml, '{}.fasta'.format(accession_version)), 'w') as fw:
+ fw.write('>{}\n{}'.format(accession_version, GBSeq_sequence.text.upper()))
- if(len(GBSeq_sequence.text) >= min_len_to_count):
- num_seq_with_len_ge_X_bp += 1
+ with open(os.path.join(dir_fasta_and_yaml, '{}.yaml'.format(accession_version)), 'w') as fw:
+ json.dump(info_for_yaml_dict, fw, indent=2)
except:
print("Unexpected error for the ID {}: {}".format(accession_version, sys.exc_info()[0]))
accession_with_errors_list.append(accession_version)