def build_index_refpkg(hrefpkg_paths, sequence_file, seqinfo, taxonomy,
dest='index.refpkg', **meta):
"""
Build an index.refpkg from a set of hrefpkgs
"""
# Clear taxonomy
taxonomy = copy.deepcopy(taxonomy)
for node in taxonomy:
node.sequence_ids = set()
def sequence_names(f):
with open(f) as fp:
r = csv.DictReader(fp)
for i in r:
yield i['seqname']
hrefpkgs = (Refpkg(i, create=False) for i in hrefpkg_paths)
seqinfo_files = (i.open_resource('seq_info') for i in hrefpkgs)
# Add seqinfo
for f in seqinfo_files:
with f:
taxonomy.populate_from_seqinfo(f)
# Remove lineages without sequences
taxonomy.prune_unrepresented()
sequence_ids = frozenset(taxonomy.subtree_sequence_ids())
with util.ntf(prefix='aln_fasta', suffix='.fasta') as tf, \
util.ntf(prefix='seq_info', suffix='.csv') as seq_info_fp, \
util.ntf(prefix='taxonomy', suffix='.csv') as tax_fp:
wrap.esl_sfetch(sequence_file, sequence_ids, tf)
tf.close()
# Seqinfo file
r = (i for i in seqinfo if i['seqname'] in sequence_ids)
w = csv.DictWriter(seq_info_fp, seqinfo[0].keys(), lineterminator='\n',
quoting=csv.QUOTE_NONNUMERIC)
w.writeheader()
w.writerows(r)
seq_info_fp.close()
taxonomy.write_taxtable(tax_fp)
tax_fp.close()
rp = Refpkg(dest, create=True)
rp.start_transaction()
rp.update_file('aln_fasta', tf.name)
rp.update_file('seq_info', seq_info_fp.name)
rp.update_file('taxonomy', tax_fp.name)
rp.update_file('profile', wrap.CM)
for k, v in meta.items():
rp.update_metadata(k, v)
rp.commit_transaction()
return rp, sequence_ids
def tax_id_refpkg(tax_id, full_tax, seqinfo, sequence_file,
output_dir='.',
index_rank='order', train_file=None, test_file=None):
"""
Build a reference package containing all descendants of tax_id from an
index reference package.
"""
with util.ntf(prefix='taxonomy', suffix='.csv') as tax_fp, \
util.ntf(prefix='aln_sto', suffix='.sto') as sto_fp, \
util.ntf(prefix='aln_fasta', suffix='.fasta') as fasta_fp, \
util.ntf(prefix='tree', suffix='.tre') as tree_fp, \
util.ntf(prefix='tree', suffix='.stats') as stats_fp, \
util.ntf(prefix='seq_info', suffix='.csv') as seq_info_fp:
# Subset taxonomy
n = full_tax.get_node(tax_id)
descendants = set(i.tax_id for i in n)
assert descendants
n.write_taxtable(tax_fp)
tax_fp.close()
# Subset seq_info
w = csv.DictWriter(seq_info_fp, seqinfo[0].keys(),
quoting=csv.QUOTE_NONNUMERIC)
w.writeheader()
rows = [i for i in seqinfo if i['tax_id'] in descendants]
sinfo = {i['seqname']: i for i in rows}
# Choose sequences, divide into train and test sets
chosen = choose_sequence_ids(n, rows, index_rank=index_rank)
keep_seq_ids = set()
train_seq_ids = set()
test_seq_ids = set()
for keep, rest in chosen:
keep_seq_ids |= frozenset(keep)
l = len(rest)
if l >= 2 * PER_TAXON:
train_seq_ids |= frozenset(rest[:l / 2])
test_seq_ids |= frozenset(rest[l / 2:])
# Picked
rows = [sinfo[i] for i in keep_seq_ids]
w.writerows(rows)
seq_info_fp.close()
# Fetch sequences
with tempfile.NamedTemporaryFile() as tf:
wrap.esl_sfetch(sequence_file,
keep_seq_ids, tf)
# Rewind
tf.seek(0)
sequences = list(SeqIO.parse(tf, 'fasta'))
logging.info("Tax id %s: %d sequences", tax_id, len(sequences))
if len(set(str(i.seq) for i in sequences)) == 1:
logging.warn("Skipping %s: only 1 unique sequence string", tax_id)
return None
# No sense in building with one sequence
if len(sequences) < 2:
logging.warn("Skipping: %d sequences.", len(sequences))
return None
# Extract training & test seqs
if train_file:
logging.info("%d training sequences", len(train_seq_ids))
wrap.esl_sfetch(sequence_file, train_seq_ids, train_file)
if test_file:
logging.info("%d test sequences", len(test_seq_ids))
wrap.esl_sfetch(sequence_file, test_seq_ids, test_file)
# Cmalign
aligned = wrap.cmalign(sequences, output=sto_fp)
aligned = list(aligned)
assert aligned
# Tree
wrap.fasttree(aligned, log_path=stats_fp.name, output_fp=tree_fp, threads=1, gtr=True)
tree_fp.close()
sto_fp.close()
SeqIO.write(aligned, fasta_fp, 'fasta')
fasta_fp.close()
rp = Refpkg(os.path.join(output_dir, tax_id + '.refpkg'), create=True)
rp.start_transaction()
rp.update_file('aln_sto', sto_fp.name)
rp.update_file('aln_fasta', fasta_fp.name)
rp.update_file('tree', tree_fp.name)
rp.update_file('seq_info', seq_info_fp.name)
rp.update_file('taxonomy', tax_fp.name)
try:
rp.update_phylo_model('FastTree', stats_fp.name)
except:
print >> sys.stderr, stats_fp.read()
raise
rp.update_file('profile', wrap.CM)
rp.commit_transaction()
util.require_executable('rppr')
rp.reroot()
return rp.path
def tax_id_refpkg(tax_id,
full_tax,
seqinfo,
sequence_file,
output_dir='.',
index_rank='order',
train_file=None,
test_file=None):
"""
Build a reference package containing all descendants of tax_id from an
index reference package.
"""
with util.ntf(prefix='taxonomy', suffix='.csv') as tax_fp, \
util.ntf(prefix='aln_sto', suffix='.sto') as sto_fp, \
util.ntf(prefix='aln_fasta', suffix='.fasta') as fasta_fp, \
util.ntf(prefix='tree', suffix='.tre') as tree_fp, \
util.ntf(prefix='tree', suffix='.stats') as stats_fp, \
util.ntf(prefix='seq_info', suffix='.csv') as seq_info_fp:
# Subset taxonomy
n = full_tax.get_node(tax_id)
descendants = set(i.tax_id for i in n)
assert descendants
n.write_taxtable(tax_fp)
tax_fp.close()
# Subset seq_info
w = csv.DictWriter(seq_info_fp,
seqinfo[0].keys(),
quoting=csv.QUOTE_NONNUMERIC)
w.writeheader()
rows = [i for i in seqinfo if i['tax_id'] in descendants]
sinfo = {i['seqname']: i for i in rows}
# Choose sequences, divide into train and test sets
chosen = choose_sequence_ids(n, rows, index_rank=index_rank)
keep_seq_ids = set()
train_seq_ids = set()
test_seq_ids = set()
for keep, rest in chosen:
keep_seq_ids |= frozenset(keep)
l = len(rest)
if l >= 2 * PER_TAXON:
train_seq_ids |= frozenset(rest[:l / 2])
test_seq_ids |= frozenset(rest[l / 2:])
# Picked
rows = [sinfo[i] for i in keep_seq_ids]
w.writerows(rows)
seq_info_fp.close()
# Fetch sequences
with tempfile.NamedTemporaryFile() as tf:
wrap.esl_sfetch(sequence_file, keep_seq_ids, tf)
# Rewind
tf.seek(0)
sequences = list(SeqIO.parse(tf, 'fasta'))
logging.info("Tax id %s: %d sequences", tax_id, len(sequences))
if len(set(str(i.seq) for i in sequences)) == 1:
logging.warn("Skipping %s: only 1 unique sequence string", tax_id)
return None
# No sense in building with one sequence
if len(sequences) < 2:
logging.warn("Skipping: %d sequences.", len(sequences))
return None
# Extract training & test seqs
if train_file:
logging.info("%d training sequences", len(train_seq_ids))
wrap.esl_sfetch(sequence_file, train_seq_ids, train_file)
if test_file:
logging.info("%d test sequences", len(test_seq_ids))
wrap.esl_sfetch(sequence_file, test_seq_ids, test_file)
# Cmalign
aligned = wrap.cmalign(sequences, output=sto_fp)
aligned = list(aligned)
assert aligned
# Tree
wrap.fasttree(aligned,
log_path=stats_fp.name,
output_fp=tree_fp,
threads=1,
gtr=True)
tree_fp.close()
sto_fp.close()
SeqIO.write(aligned, fasta_fp, 'fasta')
fasta_fp.close()
rp = Refpkg(os.path.join(output_dir, tax_id + '.refpkg'), create=True)
rp.start_transaction()
rp.update_file('aln_sto', sto_fp.name)
rp.update_file('aln_fasta', fasta_fp.name)
rp.update_file('tree', tree_fp.name)
rp.update_file('seq_info', seq_info_fp.name)
rp.update_file('taxonomy', tax_fp.name)
try:
rp.update_phylo_model('FastTree', stats_fp.name)
except:
print >> sys.stderr, stats_fp.read()
raise
rp.update_file('profile', wrap.CM)
rp.commit_transaction()
util.require_executable('rppr')
rp.reroot()
return rp.path
def build_index_refpkg(hrefpkg_paths,
sequence_file,
seqinfo,
taxonomy,
dest='index.refpkg',
**meta):
"""
Build an index.refpkg from a set of hrefpkgs
"""
# Clear taxonomy
taxonomy = copy.deepcopy(taxonomy)
for node in taxonomy:
node.sequence_ids = set()
def sequence_names(f):
with open(f) as fp:
r = csv.DictReader(fp)
for i in r:
yield i['seqname']
hrefpkgs = (Refpkg(i, create=False) for i in hrefpkg_paths)
seqinfo_files = (i.open_resource('seq_info') for i in hrefpkgs)
# Add seqinfo
for f in seqinfo_files:
with f:
taxonomy.populate_from_seqinfo(f)
# Remove lineages without sequences
taxonomy.prune_unrepresented()
sequence_ids = frozenset(taxonomy.subtree_sequence_ids())
with util.ntf(prefix='aln_fasta', suffix='.fasta') as tf, \
util.ntf(prefix='seq_info', suffix='.csv') as seq_info_fp, \
util.ntf(prefix='taxonomy', suffix='.csv') as tax_fp:
wrap.esl_sfetch(sequence_file, sequence_ids, tf)
tf.close()
# Seqinfo file
r = (i for i in seqinfo if i['seqname'] in sequence_ids)
w = csv.DictWriter(seq_info_fp,
seqinfo[0].keys(),
lineterminator='\n',
quoting=csv.QUOTE_NONNUMERIC)
w.writeheader()
w.writerows(r)
seq_info_fp.close()
taxonomy.write_taxtable(tax_fp)
tax_fp.close()
rp = Refpkg(dest, create=True)
rp.start_transaction()
rp.update_file('aln_fasta', tf.name)
rp.update_file('seq_info', seq_info_fp.name)
rp.update_file('taxonomy', tax_fp.name)
rp.update_file('profile', wrap.CM)
for k, v in meta.items():
rp.update_metadata(k, v)
rp.commit_transaction()
return rp, sequence_ids