def check_for_potential_translocation(seq, ctg_len, sorted_aligns, log_out_f):
count_ns = 0
unaligned_len = 0
prev_start = 0
for align in sorted_aligns:
# if align.start() > prev_start + 1:
if _start(align) > prev_start + 1:
# unaligned_part = seq[prev_start + 1: align.start()]
unaligned_part = seq[prev_start + 1:_start(align)]
unaligned_len += len(unaligned_part)
count_ns += unaligned_part.count('N')
# prev_start = align.end()
prev_start = _end(align)
# if ctg_len > sorted_aligns[-1].end() + 1:
if ctg_len > _end(sorted_aligns[-1]) + 1:
# unaligned_part = seq[sorted_aligns[-1].end() + 1: ctg_len]
unaligned_part = seq[_end(sorted_aligns[-1]) + 1:ctg_len]
unaligned_len += len(unaligned_part)
count_ns += unaligned_part.count('N')
# if contig consists mostly of Ns, it cannot contain interspecies translocations
if count_ns / float(
unaligned_len
) >= 0.95 or unaligned_len - count_ns < qconfig.significant_part_size:
return 0
print >> log_out_f, '\t\tIt can contain interspecies translocations.'
return 1
def intersect_and_go_next(self, align, solids, min_unique_len):
# if self.unique_end - align.end() > min_unique_len: # if enough len on the right side
if self.unique_end - _end(align) > min_unique_len: # if enough len on the right side
if self.is_solid(min_unique_len):
solids.append(self.align)
return True
self.unique_end = min(self.unique_end, _start(align) - 1)
# self.unique_end = min(self.unique_end, align.start() - 1)
return not self.is_solid(min_unique_len) # if self is not solid anymore we can switch to the next PSA
def get_added_len(set_aligns, cur_align):
last_align_idx = -2
last_align = set_aligns[last_align_idx]
# added_right = cur_align.end() - max(cur_align.start() - 1, last_align.end())
added_right = _end(cur_align) - max(_start(cur_align) - 1, _end(last_align))
added_left = 0
# while cur_align.start() < last_align.start():
while _start(cur_align) < _start(last_align):
# added_left += last_align.start() - cur_align.start()
added_left += _start(last_align) - _start(cur_align)
last_align_idx -= 1
if -last_align_idx <= len(set_aligns):
# prev_start = last_align.start() # in case of overlapping of old and new last_align
prev_start = _start(last_align) # in case of overlapping of old and new last_align
last_align = set_aligns[last_align_idx]
added_left -= max(0, min(prev_start, _end(last_align)) - _start(cur_align) + 1)
# added_left -= max(0, min(prev_start, last_align.end()) - cur_align.start() + 1)
else:
break
return added_right + added_left
def is_misassembly(align1, align2, contig_seq, ref_lens, is_cyclic=False, region_struct_variations=None):
#Calculate inconsistency between distances on the reference and on the contig
# distance_on_contig = align2.start() - align1.end() - 1
distance_on_contig = _start(align2) - _end(align1) - 1
cyclic_ref_lens = ref_lens if is_cyclic else None
if cyclic_ref_lens is not None and align1[7] == align2[7]:
distance_on_reference, cyclic_moment = distance_between_alignments(align1, align2, align1[2] < align1[3],
align2[2] < align2[3], cyclic_ref_lens[align1[7]])
else:
distance_on_reference, cyclic_moment = distance_between_alignments(align1, align2, align1[2] < align1[3],
align2[2] < align2[3])
misassembly_internal_overlap = 0
if distance_on_contig < 0:
if distance_on_reference >= 0:
misassembly_internal_overlap = (-distance_on_contig)
elif (-distance_on_reference) < (-distance_on_contig):
misassembly_internal_overlap = (distance_on_reference - distance_on_contig)
strand1 = (align1[2] <= align1[3])
strand2 = (align2[2] <= align2[3])
inconsistency = distance_on_reference - distance_on_contig
aux_data = {"inconsistency": inconsistency, "distance_on_contig": distance_on_contig,
"misassembly_internal_overlap": misassembly_internal_overlap, "cyclic_moment": cyclic_moment,
"is_sv": False, "is_scaffold_gap": False}
if qconfig.scaffolds and contig_seq and check_is_scaffold_gap(inconsistency, contig_seq, align1, align2):
aux_data["is_scaffold_gap"] = True
return False, aux_data
if region_struct_variations and check_sv(align1, align2, inconsistency, region_struct_variations):
aux_data['is_sv'] = True
return False, aux_data
if align1[7] != align2[7] and is_fragmented_ref_fake_translocation(align1, align2, ref_lens):
aux_data["inconsistency"] = sum(__get_border_gaps(align1, align2, ref_lens))
return False, aux_data
if align1[7] != align2[7] or abs(inconsistency) > qconfig.extensive_misassembly_threshold or strand1 != strand2:
return True, aux_data
return False, aux_data # regular local misassembly
def is_gap_filled_ns(contig_seq, align1, align2):
# gap_in_contig = contig_seq[align1.end(): align2.start() - 1]
gap_in_contig = contig_seq[_end(align1): _start(align2) - 1]
if len(gap_in_contig) < qconfig.Ns_break_threshold:
return False
return gap_in_contig.count('N')/len(gap_in_contig) > 0.95
def count_ns_and_not_ns_between_aligns(contig_seq, align1, align2):
# gap_in_contig = contig_seq[align1.end(): align2.start() - 1]
gap_in_contig = contig_seq[_end(align1): _start(align2) - 1]
ns_count = gap_in_contig.count('N')
return ns_count, len(gap_in_contig) - ns_count
def exclude_internal_overlaps(align1, align2, i=None, ca_output=None):
# returns size of align1[5] decrease (or 0 if not changed). It is important for cur_aligned_len calculation
def __shift_start(align, new_start, indent=''):
if ca_output is not None:
print >> ca_output.stdout_f, indent + '%s' % short_str(align),
# print >> ca_output.stdout_f, indent + '%s' % align.short_str(),
align = list(align)
if align[2] < align[3]:
align[0] += (new_start - align[2])
align[2] = new_start
align[5] = align[3] - align[2] + 1
else:
align[1] -= (new_start - align[3])
align[3] = new_start
align[5] = align[2] - align[3] + 1
align[4] = align[1] - align[0] + 1
align = tuple(align)
if ca_output is not None:
print >> ca_output.stdout_f, '--> %s' % short_str(align)
# print >> ca_output.stdout_f, '--> %s' % align.short_str()
def __shift_end(align, new_end, indent=''):
if ca_output is not None:
print >> ca_output.stdout_f, indent + '%s' % short_str(align),
# print >> ca_output.stdout_f, indent + '%s' % align.short_str(),
align = list(align)
if align[2] < align[3]:
align[1] -= (align[3] - new_end)
align[3] = new_end
align[5] = align[3] - align[2] + 1
else:
align[0] += (align[2] - new_end)
align[2] = new_end
align[5] = align[2] - align[3] + 1
align[4] = align[1] - align[0] + 1
align = tuple(align)
if ca_output is not None:
print >> ca_output.stdout_f, '--> %s' % short_str(align)
# print >> ca_output.stdout_f, '--> %s' % align.short_str()
if qconfig.ambiguity_usage == 'all':
return 0
# distance_on_contig = align2.start() - align1.end() - 1
distance_on_contig = _start(align2) - _end(align1) - 1
if distance_on_contig >= 0: # no overlap
return 0
prev_len2 = align1[5]
if ca_output is not None:
print >> ca_output.stdout_f, '\t\t\tExcluding internal overlap of size %d between Alignment %d and %d: ' \
% (-distance_on_contig, i+1, i+2),
if qconfig.ambiguity_usage == 'one': # left only one of two copies (remove overlap from shorter alignment)
if align1[5] >= align2[5]:
# __shift_start(align2, align1.end() + 1)
__shift_start(align2, _end(align1) + 1)
else:
# __shift_end(align1, align2.start() - 1)
__shift_end(align1, _start(align2) - 1)
elif qconfig.ambiguity_usage == 'none': # removing both copies
if ca_output is not None:
print >> ca_output.stdout_f
# new_end = align2.start() - 1
# __shift_start(align2, align1.end() + 1, '\t\t\t ')
new_end = _start(align2) - 1
__shift_start(align2, _end(align1) + 1, '\t\t\t ')
__shift_end(align1, new_end, '\t\t\t ')
return prev_len2 - align1[5]
def get_best_aligns_sets(sorted_aligns, ctg_len, planta_out_f, seq, ref_lens, is_cyclic=False, region_struct_variations=None):
critical_number_of_aligns = 200 # use additional optimizations for large number of alignments
penalties = dict()
penalties['extensive'] = max(50, int(round(min(qconfig.extensive_misassembly_threshold / 4.0, ctg_len * 0.05)))) - 1
penalties['local'] = max(2, int(round(min(qconfig.MAX_INDEL_LENGTH / 2.0, ctg_len * 0.01)))) - 1
penalties['scaffold'] = 5
# sorted_aligns = sorted(sorted_aligns, key=lambda x: (x.end(), x.start()))
sorted_aligns = sorted(sorted_aligns, key=lambda x: (_end(x), _start(x)))
# trying to optimise the algorithm if the number of possible alignments is large
if len(sorted_aligns) > critical_number_of_aligns:
print >> planta_out_f, '\t\t\tSkipping redundant alignments which can\'t be in the best set of alignments A PRIORI'
# FIRST STEP: find solid aligns (which are present in the best selection for sure)
# they should have unique (not covered by other aligns) region of length > 2 * extensive_penalty
min_unique_len = 2 * penalties['extensive']
possible_solids = [PSA(align) for align in sorted_aligns if align[5] > min_unique_len]
solids = []
try:
cur_PSA = possible_solids.pop()
for align in reversed(sorted_aligns):
if align != cur_PSA.align and cur_PSA.intersect_and_go_next(align, solids, min_unique_len):
next_PSA = possible_solids.pop()
while next_PSA.intersect_and_go_next(cur_PSA.align, solids, min_unique_len):
next_PSA = possible_solids.pop()
while align != next_PSA.align and next_PSA.intersect_and_go_next(align, solids, min_unique_len):
next_PSA = possible_solids.pop()
cur_PSA = next_PSA
except IndexError: # possible_solids is empty
pass
# SECOND STEP: remove all aligns which are inside solid ones
if len(solids):
solid_regions = [] # intersection of all solid aligns
cur_region = SolidRegion(solids[0])
for align in solids[1:]:
# if align.end() + 1 < cur_region.start:
if _end(align) + 1 < cur_region.start:
solid_regions.append(cur_region)
cur_region = SolidRegion(align)
else: # shift start of the current region
cur_region.start = _start(align)
# cur_region.start = align.start()
solid_regions.append(cur_region)
filtered_aligns = solids
idx = 0
try:
cur_region = solid_regions.pop()
for idx, align in enumerate(sorted_aligns):
while not cur_region.include(align):
# if align.start() > cur_region.end:
if _start(align) > cur_region.end:
cur_region = solid_regions.pop()
continue
filtered_aligns.append(align)
break
else:
print >> planta_out_f, '\t\tSkipping redundant alignment %s' % (str(align))
except IndexError: # solid_regions is empty
filtered_aligns += sorted_aligns[idx:]
# sorted_aligns = sorted(filtered_aligns, key=lambda x: (x.end(), x.start()))
sorted_aligns = sorted(filtered_aligns, key=lambda x: (_end(x), _start(x)))
# Stage 1: Dynamic programming for finding the best score
all_scored_sets = [ScoredSet(0, [], ctg_len)]
max_score = 0
for idx, align in enumerate(sorted_aligns):
local_max_score = 0
new_scored_set = None
for scored_set in all_scored_sets:
# cur_set_aligns = [sorted_aligns[i].clone() for i in scored_set.indexes] + [align.clone()]
cur_set_aligns = [_clone(sorted_aligns[i]) for i in scored_set.indexes] + [_clone(align)]
score, uncovered = get_score(scored_set.score, cur_set_aligns, ref_lens, is_cyclic, scored_set.uncovered,
seq, region_struct_variations, penalties)
if score is None: # incorrect set, i.e. internal overlap excluding resulted in incorrectly short alignment
continue
if score > local_max_score:
local_max_score = score
new_scored_set = ScoredSet(score, scored_set.indexes + [idx], uncovered)
if new_scored_set:
all_scored_sets.append(new_scored_set)
if local_max_score > max_score:
max_score = local_max_score
# Stage 2: DFS for finding multiple best sets with almost equally good score
max_allowed_score_drop = max_score - max_score * qconfig.ambiguity_score
putative_sets = [] # TODO: use priority queue -- minimal score_drop first
best_sets = []
for scored_set in all_scored_sets:
score_drop = max_score - scored_set.score
if score_drop <= max_allowed_score_drop:
heappush(putative_sets, PutativeBestSet([scored_set.indexes[-1]], score_drop, scored_set.uncovered))
ambiguity_check_is_needed = True
too_much_best_sets = False
while len(putative_sets):
putative_set = heappop(putative_sets)
# special case: no options to enlarge this set, already at the left most point
if putative_set.indexes[0] == -1:
best_sets.append(ScoredSet(max_score - putative_set.score_drop, putative_set.indexes[1:],
putative_set.uncovered))
# special case: we added the very best set and we need decide what to do next (based on ambiguity-usage)
if ambiguity_check_is_needed and len(best_sets) == 1:
if not putative_sets: # no ambiguity at all, only one good set was there
return False, too_much_best_sets, sorted_aligns, best_sets
elif not qconfig.ambiguity_usage == 'all': # several good sets are present (the contig is ambiguous) but we need only the best one
return True, too_much_best_sets, sorted_aligns, best_sets
ambiguity_check_is_needed = False
if len(best_sets) >= qconfig.BSS_MAX_SETS_NUMBER:
too_much_best_sets = (len(putative_sets) > 0)
break
continue
# the main part: trying to enlarge the set to the left (use "earlier" alignments)
align = sorted_aligns[putative_set.indexes[0]]
local_max_score = 0
local_uncovered = putative_set.uncovered
putative_predecessors = {}
for scored_set in all_scored_sets:
# we can enlarge the set with "earlier" alignments only
if scored_set.indexes and scored_set.indexes[-1] >= putative_set.indexes[0]:
break
# cur_set_aligns = [sorted_aligns[i].clone() for i in scored_set.indexes] + [align.clone()]
cur_set_aligns = [_clone(sorted_aligns[i]) for i in scored_set.indexes] + [_clone(align)]
score, uncovered = get_score(scored_set.score, cur_set_aligns, ref_lens, is_cyclic, scored_set.uncovered,
seq, region_struct_variations, penalties)
if score is not None:
putative_predecessors[scored_set] = (score, uncovered)
if score > local_max_score:
local_max_score = score
local_uncovered = uncovered
elif score == local_max_score and uncovered < local_uncovered:
local_uncovered = uncovered
for preceding_set, (score, uncovered) in putative_predecessors.iteritems():
score_drop = local_max_score - score + putative_set.score_drop
if score_drop > max_allowed_score_drop:
continue
new_index = preceding_set.indexes[-1] if preceding_set.indexes else -1
new_uncovered = uncovered + (putative_set.uncovered - local_uncovered)
heappush(putative_sets, PutativeBestSet([new_index] + putative_set.indexes,
score_drop, new_uncovered))
return True, too_much_best_sets, sorted_aligns, best_sets
def include(self, align):
return self.start <= _start(align) and _end(align) <= self.end
def __init__(self, align):
self.start = _start(align)
self.end = _end(align)
def __init__(self, align):
self.align = align
self.unique_start = _start(align)
self.unique_end = _end(align)
def analyze_contigs(ca_output,
contigs_fpath,
unaligned_fpath,
aligns,
ref_features,
ref_lens,
cyclic=None):
maxun = 10
epsilon = 0.99
umt = 0.5 # threshold for misassembled contigs with aligned less than $umt * 100% (Unaligned Missassembled Threshold)
unaligned = 0
partially_unaligned = 0
fully_unaligned_bases = 0
partially_unaligned_bases = 0
ambiguous_contigs = 0
ambiguous_contigs_extra_bases = 0
ambiguous_contigs_len = 0
partially_unaligned_with_misassembly = 0
partially_unaligned_with_significant_parts = 0
misassembly_internal_overlap = 0
contigs_with_istranslocations = 0
misassemblies_matched_sv = 0
ref_aligns = dict()
aligned_lengths = []
region_misassemblies = []
misassembled_contigs = dict()
region_struct_variations = find_all_sv(qconfig.bed)
references_misassemblies = {}
for ref in ref_labels_by_chromosomes.values():
references_misassemblies[ref] = dict(
(key, 0) for key in ref_labels_by_chromosomes.values())
# for counting SNPs and indels (both original (.all_snps) and corrected from Nucmer's local misassemblies)
total_indels_info = IndelsInfo()
unaligned_file = open(unaligned_fpath, 'w')
for contig, seq in fastaparser.read_fasta(contigs_fpath):
#Recording contig stats
ctg_len = len(seq)
print >> ca_output.stdout_f, 'CONTIG: %s (%dbp)' % (contig, ctg_len)
contig_type = 'unaligned'
#Check if this contig aligned to the reference
if contig in aligns:
for align in aligns[contig]:
#sub_seq = seq[align.start(): align.end()]
sub_seq = seq[_start(align):_end(align)]
if 'N' in sub_seq:
ns_pos = [
pos for pos in xrange(_start(align), _end(align))
if seq[pos] == 'N'
]
# ns_pos = [pos for pos in xrange(align.start(), align.end()) if seq[pos] == 'N']
contig_type = 'correct'
#Pull all aligns for this contig
num_aligns = len(aligns[contig])
#Sort aligns by aligned_length * identity - unaligned_length (as we do in BSS)
sorted_aligns = sorted(aligns[contig],
key=lambda x: (score_single_align(x), x[5]),
reverse=True)
top_len = sorted_aligns[0][5]
top_id = sorted_aligns[0][6]
top_score = score_single_align(sorted_aligns[0])
top_aligns = []
print >> ca_output.stdout_f, 'Top Length: %d Top ID: %.2f (Score: %.1f)' % (
top_len, top_id, top_score)
#Check that top hit captures most of the contig
if top_len > ctg_len * epsilon or ctg_len - top_len < maxun:
#Reset top aligns: aligns that share the same value of longest and highest identity
top_aligns.append(sorted_aligns[0])
sorted_aligns = sorted_aligns[1:]
#Continue grabbing alignments while length and identity are identical
#while sorted_aligns and top_len == sorted_aligns[0][5] and top_id == sorted_aligns[0][6]:
while sorted_aligns and (score_single_align(
sorted_aligns[0]) >=
qconfig.ambiguity_score * top_score):
top_aligns.append(sorted_aligns[0])
sorted_aligns = sorted_aligns[1:]
#Mark other alignments as insignificant (former ambiguous)
if sorted_aligns:
print >> ca_output.stdout_f, '\t\tSkipping these alignments as insignificant (option --ambiguity-score is set to "%s"):' % str(
qconfig.ambiguity_score)
for align in sorted_aligns:
print >> ca_output.stdout_f, '\t\t\tSkipping alignment ', align
if len(top_aligns) == 1:
#There is only one top align, life is good
print >> ca_output.stdout_f, '\t\tOne align captures most of this contig: %s' % str(
top_aligns[0])
# print >> ca_output.icarus_out_f, top_aligns[0].icarus_report_str()
print >> ca_output.icarus_out_f, icarus_report_str(
top_aligns[0])
ref_aligns.setdefault(top_aligns[0][7],
[]).append(top_aligns[0])
print >> ca_output.coords_filtered_f, str(top_aligns[0])
aligned_lengths.append(top_aligns[0][5])
else:
#There is more than one top align
print >> ca_output.stdout_f, '\t\tThis contig has %d significant alignments. [An ambiguously mapped contig]' % len(
top_aligns)
#Increment count of ambiguously mapped contigs and bases in them
ambiguous_contigs += 1
# we count only extra bases, so we shouldn't include bases in the first alignment
# if --ambiguity-usage is 'none', the number of extra bases will be negative!
ambiguous_contigs_len += ctg_len
# Alex: skip all alignments or count them as normal (just different aligns of one repeat). Depend on --allow-ambiguity option
if qconfig.ambiguity_usage == "none":
ambiguous_contigs_extra_bases -= top_aligns[0][5]
print >> ca_output.stdout_f, '\t\tSkipping these alignments (option --ambiguity-usage is set to "none"):'
for align in top_aligns:
print >> ca_output.stdout_f, '\t\t\tSkipping alignment ', align
elif qconfig.ambiguity_usage == "one":
ambiguous_contigs_extra_bases += 0
print >> ca_output.stdout_f, '\t\tUsing only first of these alignment (option --ambiguity-usage is set to "one"):'
print >> ca_output.stdout_f, '\t\t\tAlignment: %s' % str(
top_aligns[0])
# print >> ca_output.icarus_out_f, top_aligns[0].icarus_report_str()
print >> ca_output.icarus_out_f, icarus_report_str(
top_aligns[0])
ref_aligns.setdefault(top_aligns[0][7],
[]).append(top_aligns[0])
aligned_lengths.append(top_aligns[0][5])
print >> ca_output.coords_filtered_f, str(
top_aligns[0])
top_aligns = top_aligns[1:]
for align in top_aligns:
print >> ca_output.stdout_f, '\t\t\tSkipping alignment ', align
elif qconfig.ambiguity_usage == "all":
ambiguous_contigs_extra_bases -= top_aligns[0][5]
print >> ca_output.stdout_f, '\t\tUsing all these alignments (option --ambiguity-usage is set to "all"):'
# we count only extra bases, so we shouldn't include bases in the first alignment
first_alignment = True
while len(top_aligns):
print >> ca_output.stdout_f, '\t\t\tAlignment: %s' % str(
top_aligns[0])
# print >> ca_output.icarus_out_f, top_aligns[0].icarus_report_str(ambiguity=True)
print >> ca_output.icarus_out_f, icarus_report_str(
top_aligns[0], ambiguity=True)
ref_aligns.setdefault(top_aligns[0][7],
[]).append(top_aligns[0])
if first_alignment:
first_alignment = False
aligned_lengths.append(top_aligns[0][5])
ambiguous_contigs_extra_bases += top_aligns[0][5]
print >> ca_output.coords_filtered_f, str(
top_aligns[0]), "ambiguous"
top_aligns = top_aligns[1:]
else:
# choose appropriate alignments (to maximize total size of contig alignment and reduce # misassemblies)
is_ambiguous, too_much_best_sets, sorted_aligns, best_sets = get_best_aligns_sets(
sorted_aligns, ctg_len, ca_output.stdout_f, seq, ref_lens,
cyclic, region_struct_variations)
the_best_set = best_sets[0]
used_indexes = range(
len(sorted_aligns)
) if too_much_best_sets else get_used_indexes(best_sets)
if len(used_indexes) < len(sorted_aligns):
print >> ca_output.stdout_f, '\t\t\tSkipping redundant alignments after choosing the best set of alignments'
for idx in set(range(len(sorted_aligns))) - used_indexes:
print >> ca_output.stdout_f, '\t\tSkipping redundant alignment', sorted_aligns[
idx]
if is_ambiguous:
print >> ca_output.stdout_f, '\t\tThis contig has several significant sets of alignments. [An ambiguously mapped contig]'
# similar to regular ambiguous contigs, see above
ambiguous_contigs += 1
ambiguous_contigs_len += ctg_len
if qconfig.ambiguity_usage == "none":
ambiguous_contigs_extra_bases -= (
ctg_len - the_best_set.uncovered)
print >> ca_output.stdout_f, '\t\tSkipping all alignments in these sets (option --ambiguity-usage is set to "none"):'
for idx in used_indexes:
print >> ca_output.stdout_f, '\t\t\tSkipping alignment ', sorted_aligns[
idx]
continue
elif qconfig.ambiguity_usage == "one":
ambiguous_contigs_extra_bases += 0
print >> ca_output.stdout_f, '\t\tUsing only the very best set (option --ambiguity-usage is set to "one").'
if len(the_best_set.indexes) < len(used_indexes):
print >> ca_output.stdout_f, '\t\tSo, skipping alignments from other sets:'
for idx in used_indexes:
if idx not in the_best_set.indexes:
print >> ca_output.stdout_f, '\t\t\tSkipping alignment ', sorted_aligns[
idx]
elif qconfig.ambiguity_usage == "all":
print >> ca_output.stdout_f, '\t\tUsing all alignments in these sets (option --ambiguity-usage is set to "all"):'
print >> ca_output.stdout_f, '\t\t\tThe very best set is shown in details below, the rest are:'
for idx, cur_set in enumerate(best_sets[1:]):
print >> ca_output.stdout_f, '\t\t\t\tGroup #%d. Score: %.1f, number of alignments: %d, unaligned bases: %d' % \
(idx + 2, cur_set.score, len(cur_set.indexes), cur_set.uncovered)
if too_much_best_sets:
print >> ca_output.stdout_f, '\t\t\t\tetc...'
if len(the_best_set.indexes) < len(used_indexes):
ambiguous_contigs_extra_bases -= (
ctg_len - the_best_set.uncovered)
print >> ca_output.stdout_f, '\t\t\tList of alignments used in the sets above:'
for idx in used_indexes:
align = sorted_aligns[idx]
print >> ca_output.stdout_f, '\t\tAlignment: %s' % str(
align)
ref_aligns.setdefault(align[7],
[]).append(align)
ambiguous_contigs_extra_bases += align[5]
print >> ca_output.coords_filtered_f, str(
align), "ambiguous"
if idx not in the_best_set.indexes:
print >> ca_output.icarus_out_f, icarus_report_str(
align, is_best=False)
# print >> ca_output.icarus_out_f, align.icarus_report_str(is_best=False)
print >> ca_output.stdout_f, '\t\t\tThe best set is below. Score: %.1f, number of alignments: %d, unaligned bases: %d' % \
(the_best_set.score, len(the_best_set.indexes), the_best_set.uncovered)
real_aligns = [sorted_aligns[i] for i in the_best_set.indexes]
# main processing part
if len(real_aligns) == 1:
the_only_align = real_aligns[0]
#There is only one alignment of this contig to the reference
print >> ca_output.coords_filtered_f, str(the_only_align)
aligned_lengths.append(the_only_align[5])
# begin, end = the_only_align.start(), the_only_align.end()
begin, end = _start(the_only_align), _end(the_only_align)
unaligned_bases = 0
if (begin - 1) or (ctg_len - end):
partially_unaligned += 1
unaligned_bases = (begin - 1) + (ctg_len - end)
partially_unaligned_bases += unaligned_bases
print >> ca_output.stdout_f, '\t\tThis contig is partially unaligned. (Aligned %d out of %d bases)' % (
top_len, ctg_len)
print >> ca_output.stdout_f, '\t\tAlignment: %s' % str(
the_only_align)
# print >> ca_output.icarus_out_f, the_only_align.icarus_report_str()
print >> ca_output.icarus_out_f, icarus_report_str(
the_only_align)
if begin - 1:
print >> ca_output.stdout_f, '\t\tUnaligned bases: 1 to %d (%d)' % (
begin - 1, begin - 1)
if ctg_len - end:
print >> ca_output.stdout_f, '\t\tUnaligned bases: %d to %d (%d)' % (
end + 1, ctg_len, ctg_len - end)
# check if both parts (aligned and unaligned) have significant length
if (unaligned_bases >= qconfig.significant_part_size) and (
ctg_len - unaligned_bases >=
qconfig.significant_part_size):
print >> ca_output.stdout_f, '\t\tThis contig has both significant aligned and unaligned parts ' \
'(of length >= %d)!' % (qconfig.significant_part_size)
partially_unaligned_with_significant_parts += 1
if qconfig.meta:
contigs_with_istranslocations += check_for_potential_translocation(
seq, ctg_len, real_aligns, ca_output.stdout_f)
ref_aligns.setdefault(the_only_align[7],
[]).append(the_only_align)
else:
#Sort real alignments by position on the contig
sorted_aligns = sorted(real_aligns,
key=lambda x: (_end(x), _start(x)))
# sorted_aligns = sorted(real_aligns, key=lambda x: (x.end(), x.start()))
#There is more than one alignment of this contig to the reference
print >> ca_output.stdout_f, '\t\tThis contig is misassembled. %d total aligns.' % num_aligns
aligned_bases_in_contig = ctg_len - the_best_set.uncovered
if aligned_bases_in_contig < umt * ctg_len:
print >> ca_output.stdout_f, '\t\t\tWarning! This contig is more unaligned than misassembled. ' + \
'Contig length is %d and total length of all aligns is %d' % (ctg_len, aligned_bases_in_contig)
for align in sorted_aligns:
print >> ca_output.stdout_f, '\t\tAlignment: %s' % str(
align)
# print >> ca_output.icarus_out_f, align.icarus_report_str()
print >> ca_output.icarus_out_f, icarus_report_str(
align)
print >> ca_output.coords_filtered_f, str(align)
aligned_lengths.append(align[5])
ref_aligns.setdefault(align[7], []).append(align)
partially_unaligned_with_misassembly += 1
partially_unaligned += 1
partially_unaligned_bases += ctg_len - aligned_bases_in_contig
print >> ca_output.stdout_f, '\t\tUnaligned bases: %d' % (
ctg_len - aligned_bases_in_contig)
# check if both parts (aligned and unaligned) have significant length
if (aligned_bases_in_contig >=
qconfig.significant_part_size) and (
ctg_len - aligned_bases_in_contig >=
qconfig.significant_part_size):
print >> ca_output.stdout_f, '\t\tThis contig has both significant aligned and unaligned parts ' \
'(of length >= %d)!' % (qconfig.significant_part_size)
partially_unaligned_with_significant_parts += 1
if qconfig.meta:
contigs_with_istranslocations += check_for_potential_translocation(
seq, ctg_len, sorted_aligns,
ca_output.stdout_f)
contig_type = 'misassembled'
print >> ca_output.icarus_out_f, '\t'.join(
['CONTIG', contig,
str(ctg_len), contig_type])
print >> ca_output.stdout_f
continue
### processing misassemblies
is_misassembled, current_mio, references_misassemblies, indels_info, misassemblies_matched_sv = \
process_misassembled_contig(sorted_aligns, cyclic, aligned_lengths, region_misassemblies,
ref_lens, ref_aligns, ref_features, seq, references_misassemblies,
region_struct_variations, misassemblies_matched_sv, ca_output,
is_ambiguous)
misassembly_internal_overlap += current_mio
total_indels_info += indels_info
if is_misassembled:
misassembled_contigs[contig] = ctg_len
contig_type = 'misassembled'
if ctg_len - aligned_bases_in_contig >= qconfig.significant_part_size:
print >> ca_output.stdout_f, '\t\tThis contig has significant unaligned parts ' \
'(of length >= %d)!' % (qconfig.significant_part_size)
if qconfig.meta:
contigs_with_istranslocations += check_for_potential_translocation(
seq, ctg_len, sorted_aligns,
ca_output.stdout_f)
else:
#No aligns to this contig
print >> ca_output.stdout_f, '\t\tThis contig is unaligned. (%d bp)' % ctg_len
print >> unaligned_file, contig
#Increment unaligned contig count and bases
unaligned += 1
fully_unaligned_bases += ctg_len
print >> ca_output.stdout_f, '\t\tUnaligned bases: %d total: %d' % (
ctg_len, fully_unaligned_bases)
print >> ca_output.icarus_out_f, '\t'.join(
['CONTIG', contig, str(ctg_len), contig_type])
print >> ca_output.stdout_f
ca_output.coords_filtered_f.close()
unaligned_file.close()
misassembled_bases = sum(misassembled_contigs.itervalues())
result = {
'region_misassemblies':
region_misassemblies,
'region_struct_variations':
region_struct_variations.get_count()
if region_struct_variations else None,
'misassemblies_matched_sv':
misassemblies_matched_sv,
'misassembled_contigs':
misassembled_contigs,
'misassembled_bases':
misassembled_bases,
'misassembly_internal_overlap':
misassembly_internal_overlap,
'unaligned':
unaligned,
'partially_unaligned':
partially_unaligned,
'partially_unaligned_bases':
partially_unaligned_bases,
'fully_unaligned_bases':
fully_unaligned_bases,
'ambiguous_contigs':
ambiguous_contigs,
'ambiguous_contigs_extra_bases':
ambiguous_contigs_extra_bases,
'ambiguous_contigs_len':
ambiguous_contigs_len,
'partially_unaligned_with_misassembly':
partially_unaligned_with_misassembly,
'partially_unaligned_with_significant_parts':
partially_unaligned_with_significant_parts,
'contigs_with_istranslocations':
contigs_with_istranslocations,
'istranslocations_by_refs':
references_misassemblies
}
return result, ref_aligns, total_indels_info, aligned_lengths, misassembled_contigs