def deconcat(sequence, prev):
if prev == 'R3':
o1 = parasail.sg_qx_trace(sequence, SEQ_R5_F5, 3, 1, SCOREMAT)
o2 = parasail.sg_qx_trace(sequence, SEQ_R5_R3, 3, 1, SCOREMAT)
if o1.score >= MIN_SCORE and o1.score > o2.score:
return o1.get_traceback().comp.find(
'|'), o1.end_query, o1.score, 'F5'
elif o2.score >= MIN_SCORE:
return o2.get_traceback().comp.find(
'|'), o2.end_query, o2.score, 'R3'
else:
return None
elif prev == 'F5':
o1 = parasail.sg_qx_trace(sequence, SEQ_F3_R3, 3, 1, SCOREMAT)
o2 = parasail.sg_qx_trace(sequence, SEQ_F3_F5, 3, 1, SCOREMAT)
if o1.score >= MIN_SCORE and o1.score > o2.score:
return o1.get_traceback().comp.find(
'|'), o1.end_query, o1.score, 'R3'
elif o2.score >= MIN_SCORE:
return o2.get_traceback().comp.find(
'|'), o2.end_query, o2.score, 'F5'
else:
return None
else:
raise ValueError(
"Expected previous primer to be F5 or R3. Saw {0} instead. Abort!".
format(prev))
def find_matching_probe_pair(m1, m2, readlen, readseq, proberegions):
"""
:param m1: (shift, list of matching ext)
:param m2: (shift, list of matching lig)
:return: matched pairs if possible, otherwise whatever matches
"""
# ex: (0, [('ML_COV_675f_22PPB_00080', 'ext', 'TTTGTCACGCACTCAAAGGGA')])
shift1, list1 = m1
shift2, list2 = m2
match_candidates = []
for name1, type1, fullseq1 in list1:
if type1 != 'ext':
continue # we want the 5' to always be ext, 3' always to be lig
ext_start0, ext_end1, ext_seq = proberegions[name1]['ext']
for name2, type2, fullseq2 in list2:
if type2 != 'lig': continue
lig_start0, lig_end1, lig_seq = proberegions[name2]['lig']
expected_size = abs(lig_end1 - ext_start0)
if name1 == name2:
return ((readlen - expected_size), name1, name2, shift1,
shift2, fullseq1, fullseq2)
match_candidates.append(((readlen - expected_size), name1, name2,
shift1, shift2, fullseq1, fullseq2))
# if we reach here, we have a 5' ext, but no matching 3' lig based on the hashing, so let's try to align it
for name2, info in proberegions.items():
lig_start0, lig_end1, lig_seq = info['lig']
query_seq = readseq[-(len(lig_seq) + MAX_SW_PAD):]
o1 = parasail.sg_qx_trace(query_seq, lig_seq, 3, 1,
SW_SCORE_MATRIX)
if o1.score >= MIN_SW_SCORE:
expected_size = abs(lig_end1 - ext_start0)
return ((readlen - expected_size), name1, name1, shift1,
len(query_seq) - o1.end_query - 1, ext_seq, lig_seq)
# go through lig and find matching ext
for name2, type2, fullseq2 in list2:
if type2 != 'lig': continue
lig_start0, lig_end1, lig_seq = proberegions[name2]['lig']
for name1, info in proberegions.items():
ext_start0, ext_end1, ext_seq = info['ext']
query_seq = readseq[:(len(ext_seq) + MAX_SW_PAD)]
o1 = parasail.sg_qx_trace(query_seq, ext_seq, 3, 1,
SW_SCORE_MATRIX)
if o1.score >= MIN_SW_SCORE:
expected_size = abs(lig_end1 - ext_start0)
return ((readlen - expected_size), name2, name2,
len(query_seq) - o1.end_query - 1, shift2, ext_seq,
lig_seq)
if len(match_candidates) == 0:
return None
else:
# return match with minimal diff in expected size
match_candidates.sort(key=lambda x: abs(x[0]))
return match_candidates[0]
def trim5p3p_helper(r: SeqRecord) -> ScoreTuple:
"""
Search for 5' and 3' in the first and last 100 bp window
"""
s1 = str(r.seq[:100])
s2 = str(r.reverse_complement().seq[:100])
o1 = parasail.sg_qx_trace(s1, SEQ_5P, 3, 1, SCOREMAT)
o2 = parasail.sg_qe_db_trace(s2, SEQ_3P_REV, 3, 1, SCOREMAT)
lenA = None
if o2.score >= MINSCORE_3P:
lenA = trimA(s2[o2.end_query + 1:])
if MIN_A_LEN == 0:
end3 = len(r.seq) - o2.end_query - 1
return ScoreTuple(score5=o1.score,
end5=o1.end_query,
score3=o2.score,
end3=end3,
endA=end3)
elif lenA is not None:
end3 = len(r.seq) - o2.end_query - 1
endA = end3 - lenA + 1
return ScoreTuple(score5=o1.score,
end5=o1.end_query,
score3=o2.score,
end3=end3,
endA=endA)
else:
end3 = len(r.seq) - o2.end_query - 1
return ScoreTuple(score5=o1.score,
end5=o1.end_query,
score3=o2.score,
end3=end3,
endA=end3)
def find_probe_match_by_mapping(r, seq2, proberegions, ext_regions_by_start,
lig_regions_by_start):
i = bisect.bisect_left(ext_regions_by_start,
(r.reference_start - SW_SEARCH_PAD, '*'))
lig_last_index = min(
len(lig_regions_by_start) - 1,
bisect.bisect_right(lig_regions_by_start,
(r.reference_end + SW_SEARCH_PAD, '*'),
lo=i))
while i < len(ext_regions_by_start):
probename1 = ext_regions_by_start[i][1]
ext_start0, ext_end1, ext_seq = proberegions[probename1]['ext']
o1 = parasail.sg_qx_trace(seq2, ext_seq, 3, 1, SW_SCORE_MATRIX)
if o1.score >= MIN_SW_SCORE:
# now we need to know how long the UMI is
num1, s1 = next(iter_cigar_string(str(o1.cigar.decode, 'utf-8')))
if s1 != 'I': # don't see the UMI! ugh!
num1 = 0
j = max(0, i - MAX_SW_PAD)
while j <= lig_last_index:
probename2 = lig_regions_by_start[j][1]
lig_start0, lig_end1, lig_seq = proberegions[probename2]['lig']
o2 = parasail.sg_qx_trace(seq2[-len(lig_seq) - MAX_SW_PAD:],
lig_seq, 3, 1, SW_SCORE_MATRIX)
if o2.score >= MIN_SW_SCORE:
#pdb.set_trace()
# it's a hit of (probename1, probename2)
num2, s2 = next(
iter_cigar_string_backwards(
str(o2.cigar.decode, 'utf-8')))
if s2 != 'I':
num2 = 0
expected_size = abs(lig_end1 - ext_start0)
size_diff = len(r.seq) - num1 - num2 - expected_size
return (size_diff, probename1, probename2,\
num1, num2, ext_seq, lig_seq)
j += 1
i += 1
return None
def deconcat(sequence, prev):
if prev == "R3":
o1 = parasail.sg_qx_trace(sequence, SEQ_R5_F5, 3, 1, SCOREMAT)
o2 = parasail.sg_qx_trace(sequence, SEQ_R5_R3, 3, 1, SCOREMAT)
if o1.score >= MIN_SCORE and o1.score > o2.score:
return o1.get_traceback().comp.find("|"), o1.end_query, o1.score, "F5"
elif o2.score >= MIN_SCORE:
return o2.get_traceback().comp.find("|"), o2.end_query, o2.score, "R3"
else:
return None
elif prev == "F5":
o1 = parasail.sg_qx_trace(sequence, SEQ_F3_R3, 3, 1, SCOREMAT)
o2 = parasail.sg_qx_trace(sequence, SEQ_F3_F5, 3, 1, SCOREMAT)
if o1.score >= MIN_SCORE and o1.score > o2.score:
return o1.get_traceback().comp.find("|"), o1.end_query, o1.score, "R3"
elif o2.score >= MIN_SCORE:
return o2.get_traceback().comp.find("|"), o2.end_query, o2.score, "F5"
else:
return None
else:
raise ValueError(
f"Expected previous primer to be F5 or R3. Saw {prev} instead. Abort!"
)
def validate_reconstructed_seq(seq, orig):
"""
seq --- the sequence that is reconstructed
orig --- the original sequence
because the reconstructed seq can be longer, we don't care about deletions
(deletions w.r.t could just be exon skipping or minor base errors)
we only care that there is NOT a lot of insertions (which would indicate error in my bubble solution)
"""
o1 = parasail.sg_qx_trace(seq, orig, 3, 1, parasail.matrix_create("ACGT", 2, -5))
if o1.score < l2*2*.90: return False, o1.cigar.decode
for num, type in iter_cigar_string(o1.cigar.decode):
if type == 'I' and num > 5:
return False, o1.cigar.decode
return True, o1.cigar.decode
def sars_cov_2_s_genome(genome):
'''Returns translated SARS-CoV-2 spike sequence contained in nucleotide string'''
try:
ref_gene = str(
SeqIO.read(
f'{PACKAGE_PATH}/schemes/sars-cov-2-s/assets/MN908947.3.S.fa',
'fasta').seq)
result = parasail.sg_qx_trace(ref_gene, genome, 10, 1,
parasail.dnafull)
aln_start = result.traceback.query.rfind(ref_gene)
aln_end = aln_start + len(ref_gene)
gene = Seq(result.traceback.ref[aln_start:aln_end])
protein = str(gene.ungap().translate()).strip('*')
except Exception as e:
raise RuntimeError('Problem extracting spike sequence').with_traceback(
e.__traceback__)
return protein
def node_is_similar(seq1, seq2):
l1 = len(seq1)
l2 = len(seq2)
if l1 == 0 or l2 == 0: return False
if l1 <= 2 and l2 <= 2: return True
if l1 < l2:
l1, l2 = l2, l1
seq1, seq2 = seq2, seq1
# always make seq1 the longer one
o1 = parasail.sg_qx_trace(seq1, seq2, 3, 1, parasail.matrix_create("ACGT", 2, -5))
# require the the whole (shorter) seq2 must be aligned
# and set min score to approx 90% accuracy
if EXPECTED_ERR_RATE == 0:
return o1.score > l2*2*1.0
elif EXPECTED_ERR_RATE < 2:
return o1.score > l1*2*0.8
else:
raise Exception("Expected error rate not implemented for {0}% and above".format(EXPECTED_ERR_RATE))
return res is not None
def clip_out(bam_filename, umi_len, bc_len, output_prefix, UMI_type, shortread_bc={}, tso_len=0, g5_clip_seq=None):
"""
:param bam_filename: BAM of post-LIMA (primer-trimmed) CCS sequences
:param UMI_type: either 'A3' or 'G5' or 'G5-10X'
:param shortread_bc: a dict of barcode -> "Y|N" for top-ranked. If given, came from short read data.
--------
G5-10X
--------
5' primer -- BC --- UMI -- TSO --- GGG --- transcript --- polyA
--------
G5-clip
assumes input is like below, where the 5'/3' primer already removed by lima
Here, we will only clip out the UMI, and write out the rest of the sequence, keeping the RT + transcript
There is no assumption about the polyA tail existing or not
--------
5' primer -- UMI -- [RT primer] --- transcript --- 3' primer
"""
assert UMI_type in ('A3', 'G5', 'G5-10X', 'G5-clip')
umi_bc_len = umi_len + bc_len
if UMI_type == 'G5-clip':
try:
import parasail
except ImportError:
print("need parasail library for G5-clip mode! Abort!", file=sys.stderr)
sys.exit(-1)
para_mat = parasail.matrix_create("ACGT", 2, -5)
para_search_len = umi_len + len(g5_clip_seq) + 10
FIELDS = ['id', 'clip_len', 'extra', 'UMI', 'BC', 'BC_rev', 'BC_match', 'BC_top_rank']
if tso_len > 0: FIELDS += ['TSO']
f1 = open(output_prefix + '.trimmed.csv', 'w')
writer1 = DictWriter(f1, FIELDS, delimiter='\t', dialect='unix')
writer1.writeheader()
reader = pysam.AlignmentFile(bam_filename, 'rb', check_sq=False)
#reader = BamIO.IndexedBamReader(bam_filename)
f2 = pysam.AlignmentFile(output_prefix+'.trimmed.bam', 'wb', header=reader.header)
for r in reader:
d = r.to_dict()
#is_rev_strand = r.flag >> 4 & 1
if (r.flag >> 4 & 1):
d['seq'] = str(Seq(r.seq).reverse_complement())
d['qual'] = r.qual[::-1]
new_tags = []
for tag in d['tags']:
if tag.startswith('dq:i:') or tag.startswith('iq:i:') or tag.startswith('sq:i:'):
tag = tag[:5] + tag[::-1][:-5]
new_tags.append(tag)
d['tags'] = new_tags
d['flag'] = '4' # convert it back to not being rev complemented
if UMI_type == 'A3':
A_start, A_end = find_Aend(d['seq'])
if A_end > 0:
seq2 = d['seq'][A_end:] # should be just UMI + BC, unless UMI started with 'A's
diff = len(seq2) - umi_bc_len
if diff < 0: # UMI may have started with 'A's
seq2 = d['seq'][A_end+diff:]
seq_extra = 'NA'
if diff > 0: seq_extra = seq2[:diff]
if bc_len == 0:
seq_bc = ''
else:
seq_bc = seq2[-bc_len:]
if umi_len == 0:
seq_umi = ''
else:
if bc_len == 0:
seq_umi = seq2[-umi_len:]
else:
seq_umi = seq2[-(bc_len+umi_len):-bc_len]
# reverse complement BC because it's always listed in rev comp in short read data
seq_bc_rev = str(Seq(seq_bc).reverse_complement())
match = 'Y' if seq_bc_rev in shortread_bc else 'N'
match_top = 'Y' if (match=='Y' and shortread_bc[seq_bc_rev]=='Y') else 'N'
rec = {'id': r.qname,
'clip_len': len(seq2),
'extra': seq_extra,
'UMI': seq_umi,
'BC': seq_bc,
'BC_rev': seq_bc_rev,
'BC_match': match,
'BC_top_rank': match_top}
writer1.writerow(rec)
# subset the sequence to include only the polyAs
d['seq'] = d['seq'][:A_end]
d['qual'] = d['qual'][:A_end]
assert len(d['seq'])==len(d['qual'])
new_tags = []
for tag in d['tags']:
if tag.startswith('zs:B'): # defunct CCS tag, don't use
pass
elif tag.startswith('dq:i:') or tag.startswith('iq:i:') or tag.startswith('sq:i:'):
tag = tag[:A_end+5]
new_tags.append(tag)
else:
new_tags.append(tag)
d['tags'] = new_tags
x = pysam.AlignedSegment.from_dict(d, r.header)
f2.write(x)
elif UMI_type == 'G5':
G_start, G_end = find_Gstart(d['seq'])
if G_start > 0:
seq2 = d['seq'][:G_start] # should be just UMI
diff = len(seq2) - umi_len
if diff < 0: # UMI may have ended with Gs
seq2 = d['seq'][:G_start-diff]
seq_extra = 'NA'
if diff > 0:
seq_extra = seq2[:diff]
seq2 = seq2[diff:]
rec = {'id': r.qname,
'clip_len': len(seq2),
'extra': seq_extra,
'UMI': seq2,
'BC': 'NA', # Brendan's current design has only UMI, no BC
'BC_rev': 'NA',
'BC_match': 'NA',
'BC_top_rank': 'NA'}
writer1.writerow(rec)
# subset the sequence to remove the UMIs and "G"s
d['seq'] = d['seq'][G_end:]
d['qual'] = d['qual'][G_end:]
assert len(d['seq'])==len(d['qual'])
new_tags = []
for tag in d['tags']:
if tag.startswith('zs:B'): # defunct CCS tag, don't use
pass
elif tag.startswith('dq:i:') or tag.startswith('iq:i:') or tag.startswith('sq:i:'):
tag = tag[:5] + tag[5+G_end:]
new_tags.append(tag)
else:
new_tags.append(tag)
d['tags'] = new_tags
x = pysam.AlignedSegment.from_dict(d, r.header)
f2.write(x)
elif UMI_type == 'G5-clip':
o1 = parasail.sg_qx_trace(d['seq'][:para_search_len],
g5_clip_seq,
10,
3,
para_mat)
# 'tags': ['bx:B:i,22,20',
# ...
# 'qe:i:2835',
# 'bc:B:S,0,1',
# 'bl:Z:CCCGCGTGGCCTCCTGAATTAT',
# 'bt:Z:CATTGCCACTGTCTTCTGCT',
# 'RG:Z:70de1488/0--1']}
c_num, c_type = next(iter_cigar_string(str(o1.cigar.decode, 'utf-8')))
if c_type == 'I': # this is the (extra) + UMI
seq2 = d['seq'][:c_num]
seq_extra = 'NA'
diff = len(seq2) - umi_len
if diff < 0: # we need to get a few more bases from the primers
tag_dict = dict(x.split(':', 1) for x in d['tags'])
try:
if tag_dict['bc'] == 'B:S,0,1': # + strand
assert tag_dict['bl'].startswith('Z:')
Fseq = tag_dict['bl'][2:] # trimming away the Z:
elif tag_dict['bc'] == 'B:S,1,0': # - strand
assert tag_dict['bt'].startswith('Z:')
Fseq = str(Seq(tag_dict['bt'][2:]).reverse_complement())
seq2 = Fseq[diff:] + seq2 # rescue bases from the trimmed F primer
except KeyError:
pass # just silently not do anything and output the shorter UMI
#print("WARNING: older version of lima output, lacking 'bc' tag. Ignoring read {0}...".format(r.qname))
elif diff > 0: # there's extras
seq_extra = seq2[:diff]
seq2 = seq2[diff:]
rec = {'id': r.qname,
'clip_len': len(seq2),
'extra': seq_extra,
'UMI': seq2,
'BC': 'NA', # Brendan's current design has only UMI, no BC
'BC_rev': 'NA',
'BC_match': 'NA',
'BC_top_rank': 'NA'}
writer1.writerow(rec)
# subset the sequence to remove the UMI (but keep the G5 clip seq)
d['seq'] = d['seq'][c_num:]
d['qual'] = d['qual'][c_num:]
assert len(d['seq'])==len(d['qual'])
new_tags = []
for tag in d['tags']:
if tag.startswith('zs:B'): # defunct CCS tag, don't use
pass
elif tag.startswith('dq:i:') or tag.startswith('iq:i:') or tag.startswith('sq:i:'):
tag = tag[:5] + tag[5+c_num:]
new_tags.append(tag)
else:
new_tags.append(tag)
d['tags'] = new_tags
x = pysam.AlignedSegment.from_dict(d, r.header)
f2.write(x)
elif UMI_type == 'G5-10X':
# need to first invert the sequence so polyA is at the end
d['seq'] = str(Seq(d['seq']).reverse_complement())
d['qual'] = d['qual'][::-1]
# now it is BC -- UMI -- TSO -- GGG -- transcript -- polyA
umi_bc_tso_len = bc_len + umi_len + tso_len
G_start, G_end = find_Gstart(d['seq'][umi_bc_tso_len:umi_bc_tso_len+10])
#pdb.set_trace()
if G_start >= 0:
G_start += umi_bc_tso_len
G_end += umi_bc_tso_len
seq2 = d['seq'][:G_start] # this is BC - UMI - TSO
seq_tso = seq2[-tso_len:] + d['seq'][G_start:G_end]
diff = len(seq2) - umi_bc_tso_len
if diff > 0: # beginning may have included untrimmed primers
seq_extra = seq2[:diff]
seq2 = seq2[diff:]
seq_bc = seq2[:bc_len]
seq_umi = seq2[bc_len:umi_bc_len]
elif diff == 0:
seq_extra = 'NA'
seq_bc = seq2[:bc_len]
seq_umi = seq2[bc_len:umi_bc_len]
elif diff < 0: # we may have accidentally trimmed away some bases for BC, can't do anything
seq_extra = 'NA'
seq_bc = seq2[:bc_len+diff]
seq_umi = seq2[bc_len+diff:umi_bc_len+diff]
# reverse complement BC because it's always listed in rev comp in short read data
seq_bc_rev = str(Seq(seq_bc).reverse_complement())
match = 'Y' if seq_bc_rev in shortread_bc else 'N'
match_top = 'Y' if (match=='Y' and shortread_bc[seq_bc_rev]=='Y') else 'N'
rec = {'id': r.qname,
'clip_len': len(seq2)+(G_end-G_start),
'extra': seq_extra,
'UMI': seq_umi,
'BC': seq_bc,
'TSO': seq_tso,
'BC_rev': seq_bc_rev,
'BC_match': match,
'BC_top_rank': match_top}
writer1.writerow(rec)
# subset the sequence to remove the UMIs and "G"s
d['seq'] = d['seq'][G_end:]
d['qual'] = d['qual'][G_end:]
assert len(d['seq'])==len(d['qual'])
new_tags = []
for tag in d['tags']:
if tag.startswith('zs:B'): # defunct CCS tag, don't use
pass
elif tag.startswith('dq:i:') or tag.startswith('iq:i:') or tag.startswith('sq:i:'):
tag = tag[:5] + tag[5+G_end:]
new_tags.append(tag)
else:
new_tags.append(tag)
d['tags'] = new_tags
x = pysam.AlignedSegment.from_dict(d, r.header)
f2.write(x)
f1.close()
f2.close()
def clip_out(
bam_filename: str,
umi_len: int,
bc_len: int,
output_prefix: str,
UMI_type: umi_types,
shortread_bc: Optional[Dict[str, str]] = None,
tso_len: int = 0,
g5_clip_seq: Optional[str] = None,
) -> None:
"""
:param bam_filename: BAM of post-LIMA (primer-trimmed) CCS sequences
:param UMI_type: either 'A3' or 'G5' or 'G5-10X'
:param shortread_bc: a dict of barcode -> "Y|N" for top-ranked. If given, came from short read data.
--------
G5-10X
--------
5' primer -- BC --- UMI -- TSO --- GGG --- transcript --- polyA
--------
G5-clip
assumes input is like below, where the 5'/3' primer already removed by lima
Here, we will only clip out the UMI, and write out the rest of the sequence, keeping the RT + transcript
There is no assumption about the polyA tail existing or not
--------
5' primer -- UMI -- [RT primer] --- transcript --- 3' primer
"""
if shortread_bc is None:
shortread_bc = dict()
if UMI_type not in ("A3", "G5", "G5-10X", "G5-clip"):
raise ValueError(
f"UMI is of the wrong type. Got {UMI_type} Must be one of 'A3', 'G5', 'G5-10X', 'G5-clip'"
)
umi_bc_len = umi_len + bc_len
if UMI_type == "G5-clip":
try:
import parasail
except ImportError:
logger.error("need parasail library for G5-clip mode! Abort!")
sys.exit(-1)
para_mat = parasail.matrix_create("ACGT", 2, -5)
para_search_len = umi_len + len(g5_clip_seq) + 10
FIELDS = [
"id",
"clip_len",
"extra",
"UMI",
"BC",
"BC_rev",
"BC_match",
"BC_top_rank",
]
if tso_len > 0:
FIELDS += ["TSO"]
with pysam.AlignmentFile(bam_filename, "rb", check_sq=False) as reader:
with open(f"{output_prefix}.trimmed.csv",
"w") as f1, pysam.AlignmentFile(
f"{output_prefix}.trimmed.bam",
"wb",
header=reader.header) as f2:
writer1 = DictWriter(
f1,
FIELDS,
delimiter="\t",
dialect="unix",
)
writer1.writeheader()
for r in reader:
d = r.to_dict()
# is_rev_strand = r.flag >> 4 & 1
if r.flag >> 4 & 1:
d["seq"] = str(Seq(r.seq).reverse_complement())
d["qual"] = r.qual[::-1]
new_tags = []
for tag in d["tags"]:
if (tag.startswith("dq:i:") or tag.startswith("iq:i:")
or tag.startswith("sq:i:")):
tag = tag[:5] + tag[::-1][:-5]
new_tags.append(tag)
d["tags"] = new_tags
d["flag"] = "4" # convert it back to not being rev complemented
if UMI_type == "A3":
A_start, A_end = find_Aend(d["seq"])
if A_end > 0:
seq2 = d["seq"][
A_end:] # should be just UMI + BC, unless UMI started with 'A's
diff = len(seq2) - umi_bc_len
if diff < 0: # UMI may have started with 'A's
seq2 = d["seq"][A_end + diff:]
seq_extra = "NA"
if diff > 0:
seq_extra = seq2[:diff]
if bc_len == 0:
seq_bc = ""
else:
seq_bc = seq2[-bc_len:]
if umi_len == 0:
seq_umi = ""
else:
if bc_len == 0:
seq_umi = seq2[-umi_len:]
else:
seq_umi = seq2[-(bc_len + umi_len):-bc_len]
# reverse complement BC because it's always listed in rev comp in short read data
seq_bc_rev = str(Seq(seq_bc).reverse_complement())
match = "Y" if seq_bc_rev in shortread_bc else "N"
match_top = ("Y" if
(match == "Y"
and shortread_bc[seq_bc_rev] == "Y") else
"N")
rec = {
"id": r.qname,
"clip_len": len(seq2),
"extra": seq_extra,
"UMI": seq_umi,
"BC": seq_bc,
"BC_rev": seq_bc_rev,
"BC_match": match,
"BC_top_rank": match_top,
}
writer1.writerow(rec)
# subset the sequence to include only the polyAs
d["seq"] = d["seq"][:A_end]
d["qual"] = d["qual"][:A_end]
assert len(d["seq"]) == len(d["qual"])
new_tags = []
for tag in d["tags"]:
if tag.startswith(
"zs:B"): # defunct CCS tag, don't use
pass
elif (tag.startswith("dq:i:")
or tag.startswith("iq:i:")
or tag.startswith("sq:i:")):
tag = tag[:A_end + 5]
new_tags.append(tag)
else:
new_tags.append(tag)
d["tags"] = new_tags
x = pysam.AlignedSegment.from_dict(d, r.header)
f2.write(x)
elif UMI_type == "G5":
G_start, G_end = find_Gstart(d["seq"])
if G_start > 0:
seq2 = d["seq"][:G_start] # should be just UMI
diff = len(seq2) - umi_len
if diff < 0: # UMI may have ended with Gs
seq2 = d["seq"][:G_start - diff]
seq_extra = "NA"
if diff > 0:
seq_extra = seq2[:diff]
seq2 = seq2[diff:]
rec = {
"id": r.qname,
"clip_len": len(seq2),
"extra": seq_extra,
"UMI": seq2,
"BC":
"NA", # Brendan's current design has only UMI, no BC
"BC_rev": "NA",
"BC_match": "NA",
"BC_top_rank": "NA",
}
writer1.writerow(rec)
# subset the sequence to remove the UMIs and "G"s
d["seq"] = d["seq"][G_end:]
d["qual"] = d["qual"][G_end:]
assert len(d["seq"]) == len(d["qual"])
new_tags = []
for tag in d["tags"]:
if tag.startswith(
"zs:B"): # defunct CCS tag, don't use
pass
elif (tag.startswith("dq:i:")
or tag.startswith("iq:i:")
or tag.startswith("sq:i:")):
tag = tag[:5] + tag[5 + G_end:]
new_tags.append(tag)
else:
new_tags.append(tag)
d["tags"] = new_tags
x = pysam.AlignedSegment.from_dict(d, r.header)
f2.write(x)
elif UMI_type == "G5-clip":
o1 = parasail.sg_qx_trace(d["seq"][:para_search_len],
g5_clip_seq, 10, 3, para_mat)
# 'tags': ['bx:B:i,22,20',
# ...
# 'qe:i:2835',
# 'bc:B:S,0,1',
# 'bl:Z:CCCGCGTGGCCTCCTGAATTAT',
# 'bt:Z:CATTGCCACTGTCTTCTGCT',
# 'RG:Z:70de1488/0--1']}
c_num, c_type = next(
iter_cigar_string(str(o1.cigar.decode, "utf-8")))
if c_type == "I": # this is the (extra) + UMI
seq2 = d["seq"][:c_num]
seq_extra = "NA"
diff = len(seq2) - umi_len
if diff < 0: # we need to get a few more bases from the primers
tag_dict = dict(x.split(":", 1) for x in d["tags"])
try:
if tag_dict["bc"] == "B:S,0,1": # + strand
assert tag_dict["bl"].startswith("Z:")
Fseq = tag_dict["bl"][
2:] # trimming away the Z:
elif tag_dict["bc"] == "B:S,1,0": # - strand
assert tag_dict["bt"].startswith("Z:")
Fseq = str(
Seq(tag_dict["bt"]
[2:]).reverse_complement())
seq2 = (
Fseq[diff:] + seq2
) # rescue bases from the trimmed F primer
except KeyError:
pass # just silently not do anything and output the shorter UMI
# print("WARNING: older version of lima output, lacking 'bc' tag. Ignoring read {0}...".format(r.qname))
elif diff > 0: # there's extras
seq_extra = seq2[:diff]
seq2 = seq2[diff:]
rec = {
"id": r.qname,
"clip_len": len(seq2),
"extra": seq_extra,
"UMI": seq2,
"BC":
"NA", # Brendan's current design has only UMI, no BC
"BC_rev": "NA",
"BC_match": "NA",
"BC_top_rank": "NA",
}
writer1.writerow(rec)
# subset the sequence to remove the UMI (but keep the G5 clip seq)
d["seq"] = d["seq"][c_num:]
d["qual"] = d["qual"][c_num:]
assert len(d["seq"]) == len(d["qual"])
new_tags = []
for tag in d["tags"]:
if tag.startswith(
"zs:B"): # defunct CCS tag, don't use
pass
elif (tag.startswith("dq:i:")
or tag.startswith("iq:i:")
or tag.startswith("sq:i:")):
tag = tag[:5] + tag[5 + c_num:]
new_tags.append(tag)
else:
new_tags.append(tag)
d["tags"] = new_tags
x = pysam.AlignedSegment.from_dict(d, r.header)
f2.write(x)
elif UMI_type == "G5-10X":
# need to first invert the sequence so polyA is at the end
d["seq"] = str(Seq(d["seq"]).reverse_complement())
d["qual"] = d["qual"][::-1]
# now it is BC -- UMI -- TSO -- GGG -- transcript -- polyA
umi_bc_tso_len = bc_len + umi_len + tso_len
G_start, G_end = find_Gstart(
d["seq"][umi_bc_tso_len:umi_bc_tso_len + 10])
# pdb.set_trace()
if G_start >= 0:
G_start += umi_bc_tso_len
G_end += umi_bc_tso_len
seq2 = d["seq"][:G_start] # this is BC - UMI - TSO
seq_tso = seq2[-tso_len:] + d["seq"][G_start:G_end]
diff = len(seq2) - umi_bc_tso_len
if diff > 0: # beginning may have included untrimmed primers
seq_extra = seq2[:diff]
seq2 = seq2[diff:]
seq_bc = seq2[:bc_len]
seq_umi = seq2[bc_len:umi_bc_len]
elif diff == 0:
seq_extra = "NA"
seq_bc = seq2[:bc_len]
seq_umi = seq2[bc_len:umi_bc_len]
elif (
diff < 0
): # we may have accidentally trimmed away some bases for BC, can't do anything
seq_extra = "NA"
seq_bc = seq2[:bc_len + diff]
seq_umi = seq2[bc_len + diff:umi_bc_len + diff]
# reverse complement BC because it's always listed in rev comp in short read data
seq_bc_rev = str(Seq(seq_bc).reverse_complement())
match = "Y" if seq_bc_rev in shortread_bc else "N"
match_top = ("Y" if
(match == "Y"
and shortread_bc[seq_bc_rev] == "Y") else
"N")
rec = {
"id": r.qname,
"clip_len": len(seq2) + (G_end - G_start),
"extra": seq_extra,
"UMI": seq_umi,
"BC": seq_bc,
"TSO": seq_tso,
"BC_rev": seq_bc_rev,
"BC_match": match,
"BC_top_rank": match_top,
}
writer1.writerow(rec)
# subset the sequence to remove the UMIs and "G"s
d["seq"] = d["seq"][G_end:]
d["qual"] = d["qual"][G_end:]
assert len(d["seq"]) == len(d["qual"])
new_tags = []
for tag in d["tags"]:
if tag.startswith(
"zs:B"): # defunct CCS tag, don't use
pass
elif (tag.startswith("dq:i:")
or tag.startswith("iq:i:")
or tag.startswith("sq:i:")):
tag = tag[:5] + tag[5 + G_end:]
new_tags.append(tag)
else:
new_tags.append(tag)
d["tags"] = new_tags
x = pysam.AlignedSegment.from_dict(d, r.header)
f2.write(x)