def build_local_consensus(seqs, VERBOSE=0, store_allele_counts=False, full_cover=True):
'''Build a local consensus from an MSA
There is only ONE tricky point: what to do if some reads do not cover the whole
block, e.g. at the end of a fragment because of low coverage?
If full_cover == False, convert MSA gaps at the end of too short reads into N
Args:
seqs (list of SeqRecords): seqs to build consensus from
store_allele_counts (bool): return also allele counts from the alignment
full_cover (bool): if True, assume the reads fully cover the region (no gaps at edges)
'''
import numpy as np
from hivwholeseq.utils.miseq import alpha
from hivwholeseq.utils.mapping import align_muscle
ali = np.array(align_muscle(*seqs, sort=True), 'S1', ndmin=2)
if full_cover:
allele_counts = np.array([(ali == a).sum(axis=0) for a in alpha], int, ndmin=2)
else:
allele_counts = np.zeros((len(alpha), len(ali[0])),int)
for i in xrange(len(seqs)):
if ali[i, -1] == '-':
first_finalgap = len(ali[i].tostring().rstrip('-'))
ali[i, first_finalgap:] = 'X'
for ai, a in enumerate(alpha):
allele_counts[ai] += ali[i] == a
cov = allele_counts.sum(axis=0)
allele_counts = allele_counts[:, cov > 0]
cons_local = []
for counts in allele_counts.T:
# Pick max count nucleotide, ignoring N
maxinds = (counts[:-1] == counts.max()).nonzero()[0]
if len(maxinds) < 1:
cons_local.append('-')
continue
# Pick a random nucleotide in case of a tie
elif len(maxinds) > 1:
np.random.shuffle(maxinds)
maxind = maxinds[0]
cons_local.append(alpha[maxind])
cons_local = np.array(cons_local, 'S1')
ind_nongap = cons_local != '-'
cons_local = ''.join(cons_local[ind_nongap])
if store_allele_counts:
allele_counts = allele_counts[:, ind_nongap]
return (cons_local, allele_counts)
return cons_local
def get_overlap(data_folder, adaID, frag1, frag2, VERBOSE=0):
'''Find the overlap coordinates for the two fragments'''
from hivwholeseq.utils.mapping import align_muscle
seq1 = SeqIO.read(get_consensus_filename(data_folder, adaID, frag1), 'fasta')
seq2 = SeqIO.read(get_consensus_filename(data_folder, adaID, frag2), 'fasta')
sm1 = np.array(seq1)
sm2 = np.array(seq2)
# Find the beginning of s2 in s1
seed_len = 20
matches_min = 16
seed = sm2[:seed_len]
found = False
trials = 0
while (not found) and (trials < 3):
for pos in xrange(len(seq1) - 700, len(seq1) - seed_len):
if (sm1[pos: pos + seed_len] == seed).sum() >= matches_min - trials:
found = True
start_s2 = pos
break
if not found:
trials += 1
if not found:
return None
if VERBOSE >= 3:
print 'Beginning of '+frag2+' found in '+frag1
# In an ideal world, the overlap is a holy place in which no indels happen.
# We cannot assume that, sadly. However, we can search from the other side
# and align: find the end of s1 in s2
found = False
seed = sm1[-seed_len:]
trials = 0
while (not found) and (trials < 3):
for pos in xrange(700):
if (sm2[pos: pos + seed_len] == seed).sum() >= matches_min - trials:
found = True
end_s1 = pos + seed_len
break
if not found:
trials += 1
if not found:
return None
if VERBOSE >= 3:
print 'End of '+frag1+' found in '+frag2
# Align
ali = align_muscle(seq1[start_s2:], seq2[:end_s1])
return (start_s2, end_s1, ali)
def get_overlap(data_folder, adaID, frag1, frag2, VERBOSE=0):
"""Find the overlap coordinates for the two fragments"""
from hivwholeseq.utils.mapping import align_muscle
seq1 = SeqIO.read(get_consensus_filename(data_folder, adaID, frag1), "fasta")
seq2 = SeqIO.read(get_consensus_filename(data_folder, adaID, frag2), "fasta")
sm1 = np.array(seq1)
sm2 = np.array(seq2)
# Find the beginning of s2 in s1
seed_len = 20
matches_min = 16
seed = sm2[:seed_len]
found = False
trials = 0
while (not found) and (trials < 3):
for pos in xrange(len(seq1) - 700, len(seq1) - seed_len):
if (sm1[pos : pos + seed_len] == seed).sum() >= matches_min - trials:
found = True
start_s2 = pos
break
if not found:
trials += 1
if not found:
return None
if VERBOSE >= 3:
print "Beginning of " + frag2 + " found in " + frag1
# In an ideal world, the overlap is a holy place in which no indels happen.
# We cannot assume that, sadly. However, we can search from the other side
# and align: find the end of s1 in s2
found = False
seed = sm1[-seed_len:]
trials = 0
while (not found) and (trials < 3):
for pos in xrange(700):
if (sm2[pos : pos + seed_len] == seed).sum() >= matches_min - trials:
found = True
end_s1 = pos + seed_len
break
if not found:
trials += 1
if not found:
return None
if VERBOSE >= 3:
print "End of " + frag1 + " found in " + frag2
# Align
ali = align_muscle(seq1[start_s2:], seq2[:end_s1])
return (start_s2, end_s1, ali)
def build_msa(htseqs, VERBOSE=0):
'''Build multiple sequence alignment from cluster of haplotypes'''
from Bio.SeqRecord import SeqRecord
from Bio.Seq import Seq
from Bio.Alphabet.IUPAC import ambiguous_dna
seqs = [SeqRecord(Seq(seq, ambiguous_dna),
id='#'+str(i),
name='#'+str(i))
for i, seq in enumerate(htseqs)]
from hivwholeseq.utils.mapping import align_muscle
ali = align_muscle(*seqs, sort=True)
return ali
def build_msa(htseqs, VERBOSE=0):
'''Build multiple sequence alignment from cluster of haplotypes'''
from Bio.SeqRecord import SeqRecord
from Bio.Seq import Seq
from Bio.Alphabet.IUPAC import ambiguous_dna
seqs = [
SeqRecord(Seq(seq, ambiguous_dna), id='#' + str(i), name='#' + str(i))
for i, seq in enumerate(htseqs)
]
from hivwholeseq.utils.mapping import align_muscle
ali = align_muscle(*seqs, sort=True)
return ali
def build_msa_haplotypes(haploc, VERBOSE=0, label=''):
'''Build multiple sequence alignment from cluster of haplotypes'''
from Bio.SeqRecord import SeqRecord
from Bio.Seq import Seq
from Bio.Alphabet.IUPAC import ambiguous_dna
seqs = [SeqRecord(Seq(seq, ambiguous_dna),
id=label+'count_'+str(count)+'_rank_'+str(i),
name=label+'count_'+str(count)+'_rank_'+str(i),
description='')
for i, (seq, count) in enumerate(haploc.most_common())]
from hivwholeseq.utils.mapping import align_muscle
ali = align_muscle(*seqs, sort=True)
return ali
def write_consensus_final(seq_run, adaID, fragment, consensus):
'''Write the final consensus (fragments are now called F5 instead of F5ai)'''
dataset = MiSeq_runs[seq_run]
data_folder = dataset['folder']
samplename = dataset['samples'][dataset['adapters'].index(adaID)]
frag_out = fragment[:2]
name = samplename+'_seqrun_'+seq_run+'_adaID_'+adaID+'_'+frag_out+'_consensus'
consensusseq = SeqRecord(Seq(consensus), id=name, name=name)
outfile = get_consensus_filename(data_folder, adaID, frag_out, trim_primers=True)
SeqIO.write(consensusseq, outfile, 'fasta')
# Align all consensi via muscle and store
seqs = list(SeqIO.parse(get_reference_all_filename(data_folder, adaID, fragment), 'fasta'))
ali = align_muscle(*seqs)
AlignIO.write(ali, get_reference_all_filename(data_folder, adaID, fragment), 'fasta')
def align_consensi_dataset(dataset, adaIDs, fragments, VERBOSE=0):
'''Align consensi from different samples in a dataset'''
data_folder = dataset['folder']
# Collect consensi
if VERBOSE >= 1:
print 'Collecting consensi...',
consensi = defaultdict(dict)
for adaID in adaIDs:
samplename = dataset['samples'][dataset['adapters'].index(adaID)]
fragments_sample = samples[samplename]['fragments']
for frag in fragments_sample:
frag_gen = frag[:2]
if frag_gen not in fragments:
continue
con_fn = get_consensus_filename(data_folder, adaID, frag_gen)
if os.path.isfile(con_fn):
con = SeqIO.read(con_fn, 'fasta')
consensi[frag_gen][adaID] = con
if 'genomewide' in fragments:
frag_gens = [frag[:2] for frag in fragments_sample]
con_gw_fn = get_merged_consensus_filename(data_folder, adaID,
frag_gens)
if os.path.isfile(con_gw_fn):
con = SeqIO.read(con_gw_fn, 'fasta')
consensi['genomewide'][adaID] = con
if VERBOSE >= 1:
print 'done.'
print 'Aligning...',
# Align
alis = {}
for (frag, con_dict) in consensi.iteritems():
if VERBOSE >= 2:
print frag,
ali_frag = align_muscle(*(con_dict.values()))
alis[frag] = ali_frag
if VERBOSE >= 1:
print 'done.'
return alis
def align_consensi_dataset(dataset, adaIDs, fragments, VERBOSE=0):
'''Align consensi from different samples in a dataset'''
data_folder = dataset['folder']
# Collect consensi
if VERBOSE >= 1:
print 'Collecting consensi...',
consensi = defaultdict(dict)
for adaID in adaIDs:
samplename = dataset['samples'][dataset['adapters'].index(adaID)]
fragments_sample = samples[samplename]['fragments']
for frag in fragments_sample:
frag_gen = frag[:2]
if frag_gen not in fragments:
continue
con_fn = get_consensus_filename(data_folder, adaID, frag_gen)
if os.path.isfile(con_fn):
con = SeqIO.read(con_fn, 'fasta')
consensi[frag_gen][adaID] = con
if 'genomewide' in fragments:
frag_gens = [frag[:2] for frag in fragments_sample]
con_gw_fn = get_merged_consensus_filename(data_folder, adaID, frag_gens)
if os.path.isfile(con_gw_fn):
con = SeqIO.read(con_gw_fn, 'fasta')
consensi['genomewide'][adaID] = con
if VERBOSE >= 1:
print 'done.'
print 'Aligning...',
# Align
alis = {}
for (frag, con_dict) in consensi.iteritems():
if VERBOSE >= 2:
print frag,
ali_frag = align_muscle(*(con_dict.values()))
alis[frag] = ali_frag
if VERBOSE >= 1:
print 'done.'
return alis
def write_consensus_final(seq_run, adaID, fragment, consensus):
'''Write the final consensus (fragments are now called F5 instead of F5ai)'''
dataset = MiSeq_runs[seq_run]
data_folder = dataset['folder']
samplename = dataset['samples'][dataset['adapters'].index(adaID)]
frag_out = fragment[:2]
name = samplename + '_seqrun_' + seq_run + '_adaID_' + adaID + '_' + frag_out + '_consensus'
consensusseq = SeqRecord(Seq(consensus), id=name, name=name)
outfile = get_consensus_filename(data_folder,
adaID,
frag_out,
trim_primers=True)
SeqIO.write(consensusseq, outfile, 'fasta')
# Align all consensi via muscle and store
seqs = list(
SeqIO.parse(get_reference_all_filename(data_folder, adaID, fragment),
'fasta'))
ali = align_muscle(*seqs)
AlignIO.write(ali, get_reference_all_filename(data_folder, adaID,
fragment), 'fasta')
accept_holes=(fragment == 'genomewide'),
store_allele_counts=store_allele_counts)
if store_allele_counts:
(consensus, allele_counts) = consensus
# Store to file
if VERBOSE:
print 'Store to file'
name = samplename + '_seqrun_' + seq_run + '_adaID_' + adaID + '_' + frag_out + '_consensus'
consensusseq = SeqRecord(Seq(consensus, ambiguous_dna),
id=name,
name=name)
# Align consensus to reference via muscle and trim end gaps in ref
# (improper primer trimming in trim_and_divide)
ali = align_muscle(refseq, consensusseq, sort=True)
if ali[0][-1] == '-':
start_nongap = len(ali[0]) - len(ali[0].seq.lstrip('-'))
end_nongap = len(ali[0].seq.rstrip('-'))
ali = ali[:, start_nongap:end_nongap]
if VERBOSE >= 2:
print ali[:, :30]
print ali[:, -30:]
print 'Lenghts: ref', len(refseq), 'consensus', len(
consensusseq)
len_ali = ali.get_alignment_length()
n_diff = sum(ali[0, i] != ali[1, i] for i in xrange(len_ali))
print 'Differences from ref:', n_diff, '(' + '{:3.1f}'.format(
100.0 * n_diff / len_ali) + '%)'
block_len_initial=block_len_initial,
reads_per_alignment=n_reads_per_ali,
accept_holes=(fragment == 'genomewide'),
store_allele_counts=store_allele_counts)
if store_allele_counts:
(consensus, allele_counts) = consensus
# Store to file
if VERBOSE:
print 'Store to file'
name = samplename+'_seqrun_'+seq_run+'_adaID_'+adaID+'_'+frag_out+'_consensus'
consensusseq = SeqRecord(Seq(consensus, ambiguous_dna), id=name, name=name)
# Align consensus to reference via muscle and trim end gaps in ref
# (improper primer trimming in trim_and_divide)
ali = align_muscle(refseq, consensusseq, sort=True)
if ali[0][-1] == '-':
start_nongap = len(ali[0]) - len(ali[0].seq.lstrip('-'))
end_nongap = len(ali[0].seq.rstrip('-'))
ali = ali[:, start_nongap: end_nongap]
if VERBOSE >= 2:
print ali[:, :30]
print ali[:, -30:]
print 'Lenghts: ref', len(refseq), 'consensus', len(consensusseq)
len_ali = ali.get_alignment_length()
n_diff = sum(ali[0, i] != ali[1, i] for i in xrange(len_ali))
print 'Differences from ref:', n_diff, '('+'{:3.1f}'.format(100.0 * n_diff / len_ali)+'%)'
# Ungap consensus
cons_seq.description = ', '.join([
'Patient: ' + patient.code, time, 'region: ' + region,
'consensus'
])
cons_seq.cell_count = sample['CD4+ count']
cons_seq.viral_load = sample['viral load']
cons_seq.subtype = patient['Subtype']
seqs.append(cons_seq)
if use_joint:
seqs_all.append(cons_seq)
if VERBOSE >= 2:
print 'OK'
if VERBOSE >= 2:
print 'Align',
ali = align_muscle(*seqs, sort=True)
if VERBOSE >= 2:
print 'OK'
if use_save:
if VERBOSE >= 2:
print 'Save alignment',
fn_out = patient.get_consensi_alignment_filename(region)
mkdirs(os.path.dirname(fn_out))
AlignIO.write(ali, fn_out, 'fasta')
if VERBOSE >= 2:
print 'OK'
if VERBOSE >= 2:
print 'Build local tree'
tree = build_tree_fasttree(ali, VERBOSE=VERBOSE)
cons_seq.description = ', '.join(['Patient: '+patient.code,
time,
'region: '+region,
'consensus'])
cons_seq.cell_count = sample['CD4+ count']
cons_seq.viral_load = sample['viral load']
cons_seq.subtype = patient['Subtype']
seqs.append(cons_seq)
if use_joint:
seqs_all.append(cons_seq)
if VERBOSE >= 2:
print 'OK'
if VERBOSE >= 2:
print 'Align',
ali = align_muscle(*seqs, sort=True)
if VERBOSE >= 2:
print 'OK'
if use_save:
if VERBOSE >= 2:
print 'Save alignment',
fn_out = patient.get_consensi_alignment_filename(region)
mkdirs(os.path.dirname(fn_out))
AlignIO.write(ali, fn_out, 'fasta')
if VERBOSE >= 2:
print 'OK'
if VERBOSE >= 2:
print 'Build local tree'
tree = build_tree_fasttree(ali, VERBOSE=VERBOSE)