def cli(directory, resource, species, ver, verbose):
"""
Creates a new ensimpl snps database <filename> using Ensembl <version> and species <species>.
"""
configure_logging(verbose)
LOG = get_logger()
if ver:
ensembl_versions = list(ver)
else:
ensembl_versions = None
if species:
ensembl_species = list(species)
else:
ensembl_species = None
LOG.info("Creating database...")
tstart = time.time()
create_ensimpl_snps.create(ensembl_versions, ensembl_species, directory,
resource)
tend = time.time()
LOG.info("Creation time: {}".format(format_time(tstart, tend)))
def finalize(db, ref):
"""Finalize the database. Move everything to where it needs to be and
create the necessary indices.
Args:
db (str): Name of the database file.
ref (:obj:`ensimpl_snps.create.create_ensimpl.EnsemblReference`):
Contains information about the Ensembl reference.
"""
start = time.time()
conn = sqlite3.connect(db)
cursor = conn.cursor()
LOG.info("Finalizing database....")
sql_meta_insert = 'INSERT INTO meta_info VALUES (null, ?, ?, ?)'
meta_data = []
meta_data.append(('version', ref.version, ref.species_id))
meta_data.append(('assembly', ref.assembly, ref.species_id))
meta_data.append(('assembly_patch', ref.assembly_patch, ref.species_id))
cursor.executemany(sql_meta_insert, meta_data)
LOG.info('Creating indices...')
for sql in SQL_INDICES:
LOG.debug(sql)
cursor.execute(sql)
conn.row_factory = sqlite3.Row
LOG.info('Checking...')
for sql in SQL_SELECT_CHECKS:
LOG.debug(sql)
cursor = conn.cursor()
for row in cursor.execute(sql):
LOG.info('**** WARNING ****')
LOG.info(utils.dictify_row(cursor, row))
break
cursor.close()
LOG.info('Information')
for sql in SQL_SELECT_FINAL_INFO:
LOG.debug(sql)
cursor = conn.cursor()
for row in cursor.execute(sql):
LOG.info('{}\t{}\t{}'.format(row[0], row[1], row[2]))
cursor.close()
conn.commit()
conn.close()
LOG.info("Finalizing complete: {0}".format(
utils.format_time(start, time.time())))
def by_region(region, version, species, limit=None):
"""Perform the search by region.
Args:
region (str): The region to look for SNPs.
version (int): The Ensembl version number.
species (str): The Ensembl species identifier.
limit (int, optional): Maximum number of SNPs to return, ``None`` for
all.
Returns:
list: All the SNPs in `region`. Each element is another ``list`` with
the following values:
* chromosome
* position
* SNP identifier
* reference allele
* alternate allele
Raises:
ValueError: When `region` is empty.
"""
LOG = utils.get_logger()
LOG.debug(sqlite3.version_info)
LOG.debug(sqlite3.version)
LOG.debug('range={}'.format(region))
LOG.debug('version={}'.format(version))
LOG.debug('species_id={}'.format(species))
LOG.debug('limit={}'.format(limit))
try:
if not region:
raise ValueError('no ids were passed in')
new_region = fetch_utils.str_to_region(region)
tabix_file = fetch_utils.get_tabix_file(version, species)
tbx = pysam.TabixFile(tabix_file)
start_time = time.time()
snps = []
for row in tbx.fetch('{}'.format(new_region.chromosome),
new_region.start_position,
new_region.end_position,
parser=pysam.asTuple()):
snps.append(list(row[:5]))
LOG.info('Done: {}'.format(utils.format_time(start_time, time.time())))
return snps
except Exception as e:
LOG.error('Error: {}'.format(e))
return None
def parse_config(resource_name):
"""Take a resource string (file name, url) and open it. Parse the file.
Args:
resource_name (str): the name of the resource
Returns:
dict: :obj:`EnsemblReference` with the
key being the Ensembl version
"""
start = time.time()
all_releases = {}
line = ''
try:
with utils.open_resource(resource_name) as fd:
buffer = io.BufferedReader(fd)
line = buffer.readline() # skip first line
# parse each line and create an EnsemblReference
for line in buffer:
line = str(line, 'utf-8')
elems = line.strip().split('\t')
if len(elems) == 10:
elems.append(None)
reference = EnsemblReference(*elems)
release = all_releases.get(reference.version, {})
release[reference.species_id] = reference
all_releases[reference.version] = release
LOG.info('Config parsed in {}'.format(
utils.format_time(start, time.time())))
except IOError as io_error:
LOG.error('Unable to access resource: {}'.format(resource_name))
LOG.debug(io_error)
all_releases = None
except TypeError as type_error:
LOG.error('Unable to parse resource: {}'.format(resource_name))
LOG.debug(type_error)
LOG.debug('Error on the following:')
LOG.debug(line)
all_releases = None
return all_releases
def insert_snps(db, snps):
"""Insert snps into the database.
Args:
db (str): Name of the database file.
snps (list): A ``list`` of snps.
"""
LOG.info('Inserting snps into database: {}'.format(db))
start = time.time()
conn = sqlite3.connect(db)
sql_snps_insert = ('INSERT INTO snps ' 'VALUES (?, ?, ?, ?, ?)')
cursor = conn.cursor()
LOG.debug('Inserting {:,} snps...'.format(len(snps)))
cursor.executemany(sql_snps_insert, snps)
cursor.close()
conn.commit()
conn.close()
LOG.info('SNPs inserted in: {}'.format(
utils.format_time(start, time.time())))
def initialize(db):
"""Initialize the ensimpl_snps database.
Args:
db (str): Full path to the database file.
"""
LOG.info('Initializing database: {}'.format(db))
start = time.time()
conn = sqlite3.connect(db)
cursor = conn.cursor()
LOG.info('Generating tables...')
for sql in SQL_CREATE_TABLES:
LOG.debug(sql)
cursor.execute(sql)
cursor.close()
conn.commit()
conn.close()
LOG.info('Database initialized in: {}'.format(
utils.format_time(start, time.time())))
def cli(term, ver, exact, display, max, species, verbose):
"""
Search ensimpl database <filename> for <term>
"""
configure_logging(verbose)
LOG = get_logger()
LOG.info("Search database...")
maximum = max if max >= 0 else None
try:
tstart = time.time()
result = search_ensimpl.search(term, ver, species, exact, maximum)
tend = time.time()
LOG.debug("Num Results: {}".format(result.num_results))
count = 0
if len(result.matches) == 0:
print("No results found")
sys.exit()
headers = [
"ID", "SYMBOL", "IDS", "POSITION", "MATCH_REASON", "MATCH_VALUE"
]
tbl = []
if display in ('tab', 'csv'):
delim = '\t' if display == 'tab' else ','
print(delim.join(headers))
for match in result.matches:
line = list()
line.append(match.ensembl_gene_id)
line.append(match.symbol)
if match.external_ids:
ext_ids = []
for ids in match.external_ids:
ext_ids.append('{}/{}'.format(ids['db'], ids['db_id']))
line.append('||'.join(ext_ids))
else:
line.append('')
line.append("{}:{}-{}".format(match.chromosome,
match.position_start,
match.position_end))
line.append(match.match_reason)
line.append(match.match_value)
if display in ('tab', 'csv'):
print(delim.join(map(str, line)))
elif display == 'json':
tbl.append(dict(zip(headers, line)))
else:
tbl.append(line)
count += 1
if count >= max > 0:
break
if display in ('tab', 'csv'):
pass
elif display == 'json':
print(json.dumps({'data': tbl}, indent=4))
else:
print(tabulate(tbl, headers))
LOG.info("Search time: {}".format(format_time(tstart, tend)))
except Exception as e:
LOG.error('Error: {}'.format(e))
def by_ids(ids, version, species):
"""Perform the search for ids.
Args:
ids (list): A ``list`` of ids to look for.
version (int): The Ensembl version.
species (str): The Ensembl species identifier.
Returns:
dict: A ``dict`` withe keys return ``snps`` and ``snps_not_found``.
Raises:
ValueError: When `ids` is empty.
"""
LOG = utils.get_logger()
LOG.debug('ids={} ...'.format(ids[0:max(len(ids), 10)]))
LOG.debug('version={}'.format(version))
LOG.debug('species={}'.format(species))
try:
conn = fetch_utils.connect_to_database(version, species)
conn.row_factory = sqlite3.Row
cursor = conn.cursor()
if not ids:
raise ValueError('no ids were passed in')
temp_table = 'lookup_ids_{}'.format(utils.create_random_string())
# create a temp table and insert into
SQL_TEMP = ('CREATE TEMPORARY TABLE {} ( '
'query_id TEXT, '
'PRIMARY KEY (query_id) '
');').format(temp_table)
cursor.execute(SQL_TEMP)
SQL_TEMP = 'INSERT INTO {} VALUES (?);'.format(temp_table)
query_ids = [(_, ) for _ in ids]
cursor.executemany(SQL_TEMP, query_ids)
SQL_QUERY = ('SELECT s.* '
' FROM snps s '
' WHERE s.snp_id IN (SELECT distinct query_id FROM {}) '
' ORDER BY s.chrom, s.pos').format(temp_table)
LOG.info('Query: {}'.format(SQL_QUERY))
start_time = time.time()
snps = []
snp_ids = []
for row in cursor.execute(SQL_QUERY):
snps.append([
row['chrom'], row['pos'], row['snp_id'], row['ref'], row['alt']
])
snp_ids.append(row['snp_id'])
LOG.info('Done: {}'.format(utils.format_time(start_time, time.time())))
cursor.close()
conn.close()
snps_found = set(snp_ids)
snps_not_found = [x for x in ids if x not in snps_found]
return {'snps': snps, 'snps_not_found': snps_not_found}
except Exception as e:
LOG.error('Error: {}'.format(e))
return None