def action(args):
# remove .ssi index for sequence file if it exists
try:
os.remove(args.seqs + '.ssi')
except OSError:
pass
dtype = {'gi': str, 'tax_id': str, 'species': str}
seq_info = pd.read_csv(args.seq_info, dtype=dtype)
info_cols = seq_info.columns
if args.include:
include = args.include.read().split()
seq_info = seq_info.loc[seq_info[args.group_on].isin(include)]
# join with taxonomy if provided
if args.taxonomy:
tax = pd.read_csv(args.taxonomy, dtype=str,).set_index('tax_id')
seq_info = seq_info.join(tax, on='tax_id')
grouped = seq_info.groupby(args.group_on, sort=False)
frames = []
for key, grp in grouped:
# don't cluster groups represented by only one seq
if grp.shape[0] == 1:
clusters = mocked_cluster_output(grp['seqname'])
else:
# TODO: is longer necessarily better?
grp = grp.sort_values(
by=['is_type', 'ambig_count', 'length'],
ascending=[False, True, False])
clusters = cluster(args.seqs, grp['seqname'], identity=args.id,
prefix='{}-'.format(key), threads=args.threads)
clusters['group'] = key
frames.append(clusters)
all_clusters = pd.concat(frames)
if args.derep_map_out:
all_clusters.columns = ['type', 'seqname', 'seed', 'group']
all_clusters.to_csv(args.derep_map_out, header=True, index=False)
if args.seq_info_out:
seq_info = seq_info[seq_info['seqname'].isin(all_clusters['seed'])]
seq_info.to_csv(args.seq_info_out, columns=info_cols, index=False)
wrap.esl_sfetch(
args.seqs, all_clusters['seed'].unique(), args.seqs_out)
# finally - clean up .ssi file
os.remove(args.seqs + '.ssi')
def cluster(seqfile, seqnames, identity=1.0, prefix='cluster-', threads=None):
with util.ntf(prefix=prefix, suffix='.fasta') as fa, \
util.ntf(prefix=prefix, suffix='.uc') as uc:
wrap.esl_sfetch(seqfile, seqnames, fa)
fa.flush()
uclust.cluster(fa.name,
uc.name,
pct_id=identity,
pre_sorted=False,
quiet=True,
threads=threads)
df = uclust.parse_uclust_as_df(uc)
df = df[df.type != 'C']
df = df[['type', 'query_label', 'target_label']]
return df
def cluster(seqfile, seqnames, identity=1.0, prefix='cluster-', threads=None):
prefix = prefix.replace('/', '\\') # / confuses the filesystem
with util.ntf(prefix=prefix, suffix='.fasta') as fa, \
util.ntf(prefix=prefix, suffix='.uc') as uc:
wrap.esl_sfetch(seqfile, seqnames, fa)
fa.flush()
uclust.cluster(fa.name,
uc.name,
pct_id=identity,
pre_sorted=False,
quiet=True,
threads=threads)
df = uclust.parse_uclust_as_df(uc)
df = df[df.type != 'C']
df = df[['type', 'query_label', 'target_label']]
return df
def action(args):
# remove .ssi index for sequence file if it exists
try:
os.remove(args.seqs + '.ssi')
except OSError:
pass
dtype = {'gi': str, 'tax_id': str, 'species': str}
seq_info = pd.read_csv(args.seq_info, dtype=dtype)
info_cols = seq_info.columns
if args.include:
include = args.include.read().split()
seq_info = seq_info.loc[seq_info[args.group_on].isin(include)]
# join with taxonomy if provided
if args.taxonomy:
tax = pd.read_csv(args.taxonomy, dtype=str,).set_index('tax_id')
seq_info = seq_info.join(tax, on='tax_id')
grouped = seq_info.groupby(args.group_on, sort=False)
frames = []
for key, grp in grouped:
# don't cluster groups represented by only one seq
if grp.shape[0] == 1:
clusters = mocked_cluster_output(grp['seqname'])
else:
# TODO: is longer necessarily better?
by = []
ascending = []
for c, o in [['is_type', False],
['ambig_count', True],
['length', False]]:
if c in grp.columns:
by.append(c)
ascending.append(o)
if by:
grp = grp.sort_values(
by=by,
ascending=ascending)
clusters = cluster(
args.seqs, grp['seqname'], identity=args.id,
prefix='{}-'.format(key), threads=args.threads)
clusters['group'] = key
frames.append(clusters)
all_clusters = pd.concat(frames)
if args.derep_map_out:
all_clusters.columns = ['type', 'seqname', 'seed', 'group']
all_clusters.to_csv(args.derep_map_out, header=True, index=False)
if args.seq_info_out:
seq_info = seq_info[seq_info['seqname'].isin(all_clusters['seed'])]
seq_info.to_csv(args.seq_info_out, columns=info_cols, index=False)
wrap.esl_sfetch(
args.seqs, all_clusters['seed'].unique(), args.seqs_out)
# finally - clean up .ssi file
os.remove(args.seqs + '.ssi')
