def filter_similar_plasmids(plasmid_scores, output_dir):
"""
Fairly frequently the plasmids recovered end up being close to identical.
This method sorts through the plasmids to find which plasmids are very similar and picks the best one.
:param plasmid_scores: Dictionary generated by find_plasmid_kmer_scores.
:param output_dir: Directory to put temporary files.
:return: List of plasmids.
"""
# If only one plasmid, just return that.
if len(plasmid_scores) == 1:
return list(plasmid_scores.keys())
# Otherwise, we create a distance matrix using mash.
mash_results = list()
i = 0
for query_plasmid in plasmid_scores:
mash_results.append(list())
for reference_plasmid in plasmid_scores:
mash.dist(query_plasmid,
reference_plasmid,
output_file=os.path.join(output_dir, 'distances.tab'))
result = mash.read_mash_output(
os.path.join(output_dir, 'distances.tab'))
mash_results[i].append(result)
i += 1
matrix = list()
iteration = 1
for result in mash_results:
j = 1
for item in result:
if j > iteration:
matrix.append(item[0].distance)
j += 1
iteration += 1
# Once the distance matrix has been made, feed it into SciPy to do clustering.
z = cluster.hierarchy.linkage(matrix, method='average')
clustering = cluster.hierarchy.fcluster(z, 0.05, criterion='distance')
num_clusters = max(clustering)
clusters = list()
# Create our clusters.
plasmids_to_use = list()
for i in range(num_clusters):
clusters.append(list())
plasmid_names = list(plasmid_scores.keys())
for i in range(len(clustering)):
clusters[clustering[i] - 1].append(plasmid_names[i])
# Iterate through clusters, and use the highest scoring plasmid from each cluster for further analysis.
for group in clusters:
max_score = 0
best_hit = ''
for strain in group:
if plasmid_scores[strain] > max_score:
best_hit = strain
max_score = plasmid_scores[strain]
plasmids_to_use.append(best_hit)
return plasmids_to_use
def check_distances(ref_fasta, fastq_folder, work_dir):
bad_fastqs = list()
# fastqs = glob.glob(os.path.join(fastq_folder, '*R1*'))
mash.sketch(os.path.join(fastq_folder, '*R1*'), output_sketch=os.path.join(work_dir, 'sketch.msh'), threads=5)
mash.dist(os.path.join(work_dir, 'sketch.msh'), ref_fasta, threads=48,
output_file=os.path.join(work_dir, 'distances.tab'))
mash_output = mash.read_mash_output(os.path.join(work_dir, 'distances.tab'))
for item in mash_output:
if item.distance > 0.15: # Moved value from 0.06 to 0.15 - was definitely too conservative before.
bad_fastqs.append(item.reference)
return bad_fastqs
def check_distances(ref_fasta, fasta_folder):
bad_fastqs = list()
# fastqs = glob.glob(os.path.join(fastq_folder, '*R1*'))
mash.sketch(os.path.join(fasta_folder, '*.fasta'), output_sketch=os.path.join(fasta_folder, 'sketch.msh'), threads=56)
mash.dist(os.path.join(fasta_folder, 'sketch.msh'), ref_fasta, threads=56, output_file=os.path.join(fasta_folder, 'distances.tab'))
mash_output = mash.read_mash_output(os.path.join(fasta_folder, 'distances.tab'))
for item in mash_output:
print(item.reference, item.query, str(item.distance))
if item.distance > 0.06: # May need to adjust this value.
bad_fastqs.append(item.reference)
return bad_fastqs
def find_closest_refseq_genome(forward_reads, refseq_sketch, outdir):
print('FINDING CLOSEST REFSEQ GENOME')
out_file = os.path.join(outdir, 'distances.tab')
mash.dist(forward_reads, refseq_sketch, output_file=out_file)
closest_genome = 'NA'
closest_distance = 10000
mash_results = mash.read_mash_output(result_file=out_file)
for result in mash_results:
if result.distance < closest_distance:
print(result.distance)
print(result.query)
closest_genome = result.query
closest_distance = result.distance
return closest_genome.replace('_genomic.fna', '')
def test_mash_dist_call_kwargs():
out, err, cmd = mash.dist('tests/dummy_fastq/*fastq',
output_file='tests/distances.tab',
returncmd=True,
s='34')
assert cmd == 'mash dist tests/dummy_fastq/*fastq -p 1 -s 34 > tests/distances.tab'
os.remove('tests/distances.tab')
def test_mash_dist_call_multithreaded():
out, err, cmd = mash.dist('tests/dummy_fastq/*fastq',
output_file='tests/distances.tab',
returncmd=True,
threads=4)
assert cmd == 'mash dist tests/dummy_fastq/*fastq -p 4 > tests/distances.tab'
os.remove('tests/distances.tab')
def test_read_mash_dist():
out, err, cmd = mash.dist('tests/dummy_fastq/*fastq',
output_file='tests/distances.tab',
returncmd=True)
results = mash.read_mash_output('tests/distances.tab')
assert results[1].reference == 'tests/dummy_fastq/single.fastq' \
and results[1].query == 'tests/dummy_fastq/test_R2.fastq' \
and results[1].distance == 0.00763536
os.remove('tests/distances.tab')
def closerelatives_redmine(redmine_instance, issue, work_dir, description):
sentry_sdk.init(SENTRY_DSN, before_send=before_send)
# Unpickle Redmine objects
redmine_instance = pickle.load(open(redmine_instance, 'rb'))
issue = pickle.load(open(issue, 'rb'))
description = pickle.load(open(description, 'rb'))
try:
# First line of description should be number of close relatives desired.
try:
num_close_relatives = int(description[0])
except ValueError:
redmine_instance.issue.update(resource_id=issue.id,
notes='Error! The first line of the description must be the number'
' of strains you want to find. The first line of your '
'description was: {}'.format(description[0]),
status_id=4)
return
# Second line of description should be the SEQID of what you want to find a close reference for.
seqid = description[1]
# Try to extract FASTA files for the specified SEQID.
retrieve_nas_files(seqids=[seqid],
outdir=os.path.join(work_dir, 'fasta'),
filetype='fasta',
copyflag=False)
if len(glob.glob(os.path.join(work_dir, 'fasta', '*.fasta'))) != 1:
redmine_instance.issue.update(resource_id=issue.id,
notes='Error! Could not find FASTA file for the specified SEQID. The SEQID'
' that you specified was: {}'.format(seqid),
status_id=4)
return
# Run mash dist with the FASTQ file specified against the sketch of all our stuff.
query_fasta = glob.glob(os.path.join(work_dir, 'fasta', '*.fasta'))[0]
mash.dist(query_fasta, '/mnt/nas2/redmine/bio_requests/14674/all_sequences.msh',
threads=8, output_file=os.path.join(work_dir, 'distances.tab'))
mash_results = mash.read_mash_output(os.path.join(work_dir, 'distances.tab'))
result_dict = dict()
# Put all the results into a dictionary, where the key is the sequence file and the value is mash distance
# between query fastq and reference fastq.
for item in mash_results:
seq_name = os.path.split(item.query)[-1].split('_')[0]
result_dict[seq_name] = item.distance
# Sort the results, store the sorted dictionary keys in a list.
sorted_distance_results = sorted(result_dict, key=result_dict.get)
# Prepare a string that lists the top hit SEQIDs to be posted to redmine.
upload_string = ''
for i in range(num_close_relatives):
upload_string = upload_string + sorted_distance_results[i].replace('.fasta', '') + ' (' + str(result_dict[sorted_distance_results[i]]) + ')\n'
# Also make a CSV file of all results, in case someone wants to take a closer look.
with open(os.path.join(work_dir, 'close_relatives_results.csv'), 'w') as f:
f.write('Strain,MashDistance\n')
for seq in sorted_distance_results:
f.write('{},{}\n'.format(seq.replace('.fasta', ''), result_dict[seq]))
output_list = [
{
'path': os.path.join(work_dir, 'close_relatives_results.csv'),
'filename': 'close_relatives_results.csv'
}
]
# Post the list of closely related SEQIDs to redmine, as well as the CSV result file.
redmine_instance.issue.update(resource_id=issue.id,
notes='Process complete! Here is the list of the {num_relatives} closest strains '
'to {query_strain} (mash distance between query and result in brackets):'
'\n{upload_string}'.format(num_relatives=str(num_close_relatives),
query_strain=seqid,
upload_string=upload_string),
status_id=4,
uploads=output_list)
except Exception as e:
sentry_sdk.capture_exception(e)
redmine_instance.issue.update(resource_id=issue.id,
notes='Something went wrong! We log this automatically and will look into the '
'problem and get back to you with a fix soon.')
def test_dist_no_input_files():
with pytest.raises(ValueError):
mash.dist()
