def predict(self):
self.h2_prediction = self.get_antigen_gene_blast_results(
self.h2_prediction, FLJB_FASTA_PATH)
if not self.h2_prediction.is_missing:
if not self.h2_prediction.is_perfect_match:
top_result = self.h2_prediction.top_result
match_len = top_result['length']
pident = top_result['pident']
# short lower %ID matches are treated as missing or '-' for H2
if match_len <= 600 and pident < 88.0:
self.h2_prediction.h2 = '-'
self.h2_prediction.is_missing = True
return
if match_len <= 600 and not self.h2_prediction.is_trunc:
self.h2_prediction.h2 = '-'
self.h2_prediction.is_missing = True
return
df_blast_results = pd.DataFrame(
self.h2_prediction.blast_results)
df_blast_results = df_blast_results[
(df_blast_results['mismatch'] <= 50)
& (df_blast_results['length'] >= 700)]
if df_blast_results.shape[0] == 0:
self.h2_prediction.is_missing = True
self.h2_prediction.top_result = None
self.h2_prediction.h2 = '-'
return
df_blast_results_over1000 = df_blast_results[
(df_blast_results['mismatch'] <= 5)
& (df_blast_results['length'] >= 1000)]
if df_blast_results_over1000.shape[0] > 0:
df_blast_results = df_blast_results_over1000.sort_values(
by='mismatch')
else:
df_blast_results = df_blast_results.sort_values(
by='bitscore', ascending=False)
result_dict = BlastReader.df_first_row_to_dict(
df_blast_results)
result_trunc = BlastReader.is_blast_result_trunc(
qstart=result_dict['qstart'],
qend=result_dict['qend'],
sstart=result_dict['sstart'],
send=result_dict['send'],
qlen=result_dict['qlen'],
slen=result_dict['slen'])
self.h2_prediction.top_result = result_dict
self.h2_prediction.is_trunc = result_trunc
self.h2_prediction.h2 = get_antigen_name(
self.h2_prediction.top_result['qseqid'])
if self.h2_prediction.is_missing:
self.h2_prediction.h2 = '-'
def get_antigen_gene_blast_results(self, model_obj, antigen_gene_fasta):
blast_outfile = self.blast_runner.blast_against_query(
antigen_gene_fasta)
blast_reader = BlastReader(blast_outfile)
is_missing = blast_reader.is_missing
model_obj.is_missing = is_missing
if not is_missing:
model_obj.blast_results = blast_reader.df_dict()
model_obj.top_result = blast_reader.top_result()
model_obj.is_perfect_match = blast_reader.is_perfect_match
model_obj.is_trunc = blast_reader.is_trunc
return model_obj
def predict(self, filter=['N/A']):
self.h1_prediction = self.get_antigen_gene_blast_results(
self.h1_prediction, FLIC_FASTA_PATH, filter)
if not self.h1_prediction.is_missing and self.h1_prediction.top_result is not None:
if not self.h1_prediction.is_perfect_match:
df_blast_results = pd.DataFrame(
self.h1_prediction.blast_results)
df_blast_results = df_blast_results[
(df_blast_results['mismatch'] <= 25)
& (df_blast_results['length'] >= 700)]
if df_blast_results.shape[0] == 0:
df_blast_results = pd.DataFrame(
self.h1_prediction.blast_results)
df_blast_results = df_blast_results[
(df_blast_results['mismatch'] <= 0)
& (df_blast_results['length'] >= 400)]
if df_blast_results.shape[0] == 0:
self.h1_prediction.is_missing = True
self.h1_prediction.top_result = None
self.h1_prediction.h1 = None
return
df_blast_results_over1000 = df_blast_results[
(df_blast_results['mismatch'] <= 5)
& (df_blast_results['length'] >= 1000)]
if df_blast_results_over1000.shape[0] > 0:
df_blast_results = df_blast_results_over1000.sort_values(
by='mismatch')
else:
df_blast_results = df_blast_results.sort_values(
by='bitscore', ascending=False)
result_dict = BlastReader.df_first_row_to_dict(
df_blast_results)
result_trunc = BlastReader.is_blast_result_trunc(
qstart=result_dict['qstart'],
qend=result_dict['qend'],
sstart=result_dict['sstart'],
send=result_dict['send'],
qlen=result_dict['qlen'],
slen=result_dict['slen'])
self.h1_prediction.top_result = result_dict
self.h1_prediction.is_trunc = result_trunc
self.h1_prediction.h1 = get_antigen_name(
self.h1_prediction.top_result['qseqid'])
def test_BlastReader_is_blast_result_trunc():
# not truncated; match found in the middle of the subject sequence
assert not BlastReader.is_blast_result_trunc(qstart=1,
qend=100,
sstart=101,
send=200,
qlen=100,
slen=1000)
# not truncated; shorter match (-10bp) found in the middle of the subject
# sequence
assert not BlastReader.is_blast_result_trunc(qstart=1,
qend=90,
sstart=101,
send=190,
qlen=100,
slen=1000)
# not truncated; shorter match (-20bp) found in the middle of the subject
# sequence
assert not BlastReader.is_blast_result_trunc(qstart=1,
qend=80,
sstart=101,
send=180,
qlen=100,
slen=1000)
# truncated at the start of the subject
assert BlastReader.is_blast_result_trunc(qstart=51,
qend=100,
sstart=1,
send=50,
qlen=100,
slen=1000)
# truncated at the end of the subject
assert BlastReader.is_blast_result_trunc(qstart=51,
qend=100,
sstart=951,
send=1000,
qlen=100,
slen=1000)
def run_cgmlst(blast_runner, full=False):
"""Perform in silico cgMLST on an input genome
Args:
blast_runner (sistr.src.blast_wrapper.BlastRunner): blastn runner object with genome fasta initialized
Returns:
dict: cgMLST ref genome match, distance to closest ref genome, subspecies and serovar predictions
dict: marker allele match results (seq, allele name, blastn results)
"""
from sistr.src.serovar_prediction.constants import genomes_to_serovar
df_cgmlst_profiles = ref_cgmlst_profiles()
logging.debug('{} distinct cgMLST330 profiles'.format(
df_cgmlst_profiles.shape[0]))
logging.info('Running BLAST on serovar predictive cgMLST330 alleles')
cgmlst_fasta_path = CGMLST_CENTROID_FASTA_PATH if not full else CGMLST_FULL_FASTA_PATH
blast_outfile = blast_runner.blast_against_query(cgmlst_fasta_path)
logging.info('Reading BLAST output file "{}"'.format(blast_outfile))
blast_reader = BlastReader(blast_outfile)
if blast_reader.df is None:
logging.error('No cgMLST330 alleles found!')
return (
{
'distance': 1.0,
'genome_match': None,
'serovar': None,
'matching_alleles': 0,
'subspecies': None,
'cgmlst330_ST': None,
},
{},
)
logging.info('Found {} cgMLST330 allele BLAST results'.format(
blast_reader.df.shape[0]))
df_cgmlst_blastn = process_cgmlst_results(blast_reader.df)
marker_match_results = matches_to_marker_results(
df_cgmlst_blastn[df_cgmlst_blastn.is_match])
contig_blastn_records = alleles_to_retrieve(df_cgmlst_blastn)
retrieved_marker_alleles = get_allele_sequences(blast_runner.fasta_path,
contig_blastn_records,
full=full)
logging.info('Type retrieved_marker_alleles %s',
type(retrieved_marker_alleles))
all_marker_results = marker_match_results.copy()
found_cgmlst_genes = 0
for marker, res in retrieved_marker_alleles.items():
all_marker_results[marker] = res
for marker in df_cgmlst_profiles.columns:
if marker not in all_marker_results:
all_marker_results[marker] = {
'blast_result': None,
'name': None,
'seq': None,
}
cgmlst_results = {}
for marker, res in all_marker_results.items():
try:
cgmlst_results[marker] = int(res['name'])
found_cgmlst_genes += 1
except:
logging.error('Missing cgmlst_results for %s', marker)
logging.debug(res)
logging.info(
'Calculating number of matching alleles to serovar predictive cgMLST330 profiles'
)
df_relatives = find_closest_related_genome(cgmlst_results,
df_cgmlst_profiles)
genome_serovar_dict = genomes_to_serovar()
df_relatives['serovar'] = [
genome_serovar_dict[genome] for genome in df_relatives.index
]
logging.debug('Top 5 serovar predictive cgMLST profiles:\n{}'.format(
df_relatives.head()))
spp = None
subspeciation_tuple = cgmlst_subspecies_call(df_relatives)
if subspeciation_tuple is not None:
spp, distance, spp_counter = subspeciation_tuple
logging.info(
'Top subspecies by cgMLST is "{}" (min dist={}, Counter={})'.
format(spp, distance, spp_counter))
else:
logging.warning('Subspeciation by cgMLST was not possible!')
cgmlst_serovar = None
cgmlst_matching_genome = None
cgmlst_matching_alleles = 0
cgmlst_distance = 1.0
for idx, row in df_relatives.iterrows():
cgmlst_distance = row['distance']
cgmlst_matching_alleles = row['matching']
cgmlst_found_loci = found_cgmlst_genes
cgmlst_serovar = row['serovar'] if cgmlst_distance <= 1.0 else None
cgmlst_matching_genome = idx if cgmlst_distance <= 1.0 else None
logging.info(
'Top serovar by cgMLST profile matching: "{}" with {} matching alleles, distance={:.1%}'
.format(cgmlst_serovar, cgmlst_matching_alleles, cgmlst_distance))
break
cgmlst_st = None
cgmlst_markers_sorted = sorted(all_marker_results.keys())
cgmlst_allele_names = []
marker = None
for marker in cgmlst_markers_sorted:
try:
aname = all_marker_results[marker]['name']
if aname:
cgmlst_allele_names.append(str(aname))
else:
break
except:
break
if len(cgmlst_allele_names) == len(cgmlst_markers_sorted):
cgmlst_st = allele_name('-'.join(cgmlst_allele_names))
logging.info('cgMLST330 Sequence Type=%s', cgmlst_st)
else:
logging.warning(
'Could not compute cgMLST330 Sequence Type due to missing data (marker %s)',
marker)
return (
{
'distance': cgmlst_distance,
'genome_match': cgmlst_matching_genome,
'serovar': cgmlst_serovar,
'matching_alleles': cgmlst_matching_alleles,
'found_loci': cgmlst_found_loci,
'subspecies': spp,
'cgmlst330_ST': cgmlst_st,
},
all_marker_results,
)
def test_BlastReader(blast_runner):
blast_outfile = blast_runner.run_blast(WZX_FASTA_PATH)
blast_reader = BlastReader(blast_outfile)
top_result = blast_reader.top_result()
assert get_antigen_name(top_result['qseqid']) == 'O58'