def execute_alignment(self, query):
row = []
self.dna_alphabet = "AGTCNRYSWKMBDHV"
matrix = ssw.DNA_ScoreMatrix(alphabet=self.dna_alphabet)
aligner = ssw.Aligner(matrix=matrix)
for reference in self.references:
alignment = aligner.align(query, reference)
row.append(alignment)
row.sort(cmp=lambda x, y: cmp(x.score, y.score), reverse=True)
winner = row[0]
return (winner.reference, winner)
def get_ssw_alignments(best_edit_distances, querys, targets):
score_matrix = ssw.DNA_ScoreMatrix(match=1, mismatch=-2)
aligner = ssw.Aligner(gap_open=2, gap_extend=1, matrix=score_matrix)
best_edit_distances_ssw = {}
best_cigars_ssw = {}
for acc1 in best_edit_distances:
seq1 = querys[acc1]
best_ed = len(seq1)
best_edit_distances_ssw[acc1] = {}
best_cigars_ssw[acc1] = {}
for acc2 in best_edit_distances[acc1]:
seq2 = targets[acc2]
result = aligner.align(seq1, seq2, revcomp=False)
seq2_aln, match_line, seq1_aln = result.alignment
matches, mismatches, indels = match_line.count(
"|"), match_line.count("*"), match_line.count(" ")
insertion_count = seq2_aln.count("-")
deletion_count = seq1_aln.count("-")
sw_ed = mismatches + indels
best_edit_distances_ssw[acc1][
acc2] = sw_ed # (deletion_count, insertion_count, mismatches )
seq1_aln, match_line, seq2_aln = result.alignment
best_cigars_ssw[acc1][acc2] = (result.cigar, mismatches, indels,
result.query_begin,
len(seq1) - result.query_end - 1,
result.reference_begin, len(seq2) -
result.reference_end - 1)
# print(acc1,acc2)
# print(result.query_begin, len(seq1) - result.query_end - 1, result.reference_begin, len(seq2) - result.reference_end -1, result.cigar, mismatches, indels)
# print()
# print(sw_ed, (deletion_count, insertion_count, mismatches ))
# print(seq1_aln)
# print(match_line)
# print(seq2_aln)
# edit_distance, locations, cigar = edlib_traceback(seq1, seq2, k =1000)
# print(edit_distance, locations, cigar)
# print()
# for acc in best_cigars_ssw:
# if len(best_cigars_ssw[acc]) ==0:
# print("!!!!", acc)
# print(len(best_cigars_ssw))
# sys.exit()
return best_edit_distances_ssw, best_cigars_ssw
def ssw_alignment( x_acc, y_acc, x, y, i,j, max_discrepancy = 50):
"""
Aligns two sequences with SSW
x: query
y: reference
"""
if i % 10 == 0 and j % 10000 == 0:
print("processing alignments on all y's where read x_i is participating. i={0}".format(i+1))
score_matrix = ssw.DNA_ScoreMatrix(match=1, mismatch=-1)
aligner = ssw.Aligner(gap_open=2, gap_extend=1, matrix=score_matrix)
# for the ends that SSW leaves behind
bio_matrix = matlist.blosum62
g_open = -1
g_extend = -0.5
######################################
result = aligner.align(x, y, revcomp=True)
y_alignment, match_line, x_alignment = result.alignment
matches, mismatches, indels = match_line.count("|"), match_line.count("*"), match_line.count(" ")
deletions = x_alignment.count("-")
insertions = y_alignment.count("-")
assert deletions + insertions == indels
# alignment_length = len(match_line)
start_discrepancy = max(result.query_begin, result.reference_begin) # 0-indexed # max(result.query_begin, result.reference_begin) - min(result.query_begin, result.reference_begin)
query_end_discrepancy = len(x) - result.query_end - 1
ref_end_discrepancy = len(y) - result.reference_end - 1
end_discrepancy = max(query_end_discrepancy, ref_end_discrepancy) # max(result.query_end, result.reference_end) - min(result.query_end, result.reference_end)
# print(query_end_discrepancy, ref_end_discrepancy)
tot_discrepancy = start_discrepancy + end_discrepancy
if 0 < start_discrepancy <= max_discrepancy:
# print("HERE")
matches_snippet = 0
mismatches_snippet = 0
if result.query_begin and result.reference_begin:
query_start_snippet = x[:result.query_begin]
ref_start_snippet = y[:result.reference_begin]
alns = pairwise2.align.globalds(query_start_snippet, ref_start_snippet, bio_matrix, g_open, g_extend)
top_aln = alns[0]
# print(alns)
mismatches_snippet = len(list(filter(lambda x: x[0] != x[1] and x[0] != '-' and x[1] != "-", zip(top_aln[0],top_aln[1]))))
indels_snippet = top_aln[0].count("-") + top_aln[1].count("-")
matches_snippet = len(top_aln[0]) - mismatches_snippet - indels_snippet
# print(matches_snippet, mismatches_snippet, indels_snippet)
query_start_alignment_snippet = top_aln[0]
ref_start_alignment_snippet = top_aln[1]
elif result.query_begin:
query_start_alignment_snippet = x[:result.query_begin]
ref_start_alignment_snippet = "-"*len(query_start_alignment_snippet)
indels_snippet = len(ref_start_alignment_snippet)
elif result.reference_begin:
ref_start_alignment_snippet = y[:result.reference_begin]
query_start_alignment_snippet = "-"*len(ref_start_alignment_snippet)
indels_snippet = len(query_start_alignment_snippet)
else:
print("BUG")
sys.exit()
matches, mismatches, indels = matches + matches_snippet, mismatches + mismatches_snippet, indels + indels_snippet
# print(ref_start_alignment_snippet)
# print(query_start_alignment_snippet)
y_alignment = ref_start_alignment_snippet + y_alignment
x_alignment = query_start_alignment_snippet + x_alignment
if 0 < end_discrepancy <= max_discrepancy:
# print("HERE2", query_end_discrepancy, ref_end_discrepancy)
# print(y_alignment)
# print(y)
# print(match_line)
# print(x_alignment)
# print(x)
# print(matches, len(x_alignment))
matches_snippet = 0
mismatches_snippet = 0
if query_end_discrepancy and ref_end_discrepancy:
query_end_snippet = x[result.query_end+1:]
ref_end_snippet = y[result.reference_end+1:]
alns = pairwise2.align.globalds(query_end_snippet, ref_end_snippet, bio_matrix, g_open, g_extend)
top_aln = alns[0]
mismatches_snippet = len(list(filter(lambda x: x[0] != x[1] and x[0] != '-' and x[1] != "-", zip(top_aln[0],top_aln[1]))))
indels_snippet = top_aln[0].count("-") + top_aln[1].count("-")
matches_snippet = len(top_aln[0]) - mismatches_snippet - indels_snippet
query_end_alignment_snippet = top_aln[0]
ref_end_alignment_snippet = top_aln[1]
elif query_end_discrepancy:
query_end_alignment_snippet = x[result.query_end+1:]
ref_end_alignment_snippet = "-"*len(query_end_alignment_snippet)
indels_snippet = len(ref_end_alignment_snippet)
elif ref_end_discrepancy:
ref_end_alignment_snippet = y[result.reference_end+1:]
query_end_alignment_snippet = "-"*len(ref_end_alignment_snippet)
indels_snippet = len(query_end_alignment_snippet)
else:
print("BUG")
sys.exit()
matches, mismatches, indels = matches + matches_snippet, mismatches + mismatches_snippet, indels + indels_snippet
y_alignment = y_alignment + ref_end_alignment_snippet
x_alignment = x_alignment + query_end_alignment_snippet
# matches, mismatches, indels = match_line.count("|"), match_line.count("*"), match_line.count(" ")
deletions = x_alignment.count("-")
insertions = y_alignment.count("-")
assert deletions + insertions == indels
if start_discrepancy > max_discrepancy or end_discrepancy > max_discrepancy:
# print("REMOVING", start_discrepancy, end_discrepancy)
return (y_alignment, x_alignment, None)
else:
return (y_alignment, x_alignment, (matches, mismatches, indels, deletions, insertions))
def ssw_alignment(x, y, ends_discrepancy_threshold = 250 ):
"""
Aligns two sequences with SSW
x: query
y: reference
"""
score_matrix = ssw.DNA_ScoreMatrix(match=1, mismatch=-20)
aligner = ssw.Aligner(gap_open=50, gap_extend=0, matrix=score_matrix)
# for the ends that SSW leaves behind
bio_matrix = matlist.blosum62
g_open = -1
g_extend = -0.5
######################################
# result = aligner.align("GA", "G", revcomp=False)
# y_alignment, match_line, x_alignment = result.alignment
# c = Counter(match_line)
# matches, mismatches, indels = c["|"], c["*"], c[" "]
# alignment_length = len(match_line)
# print("matches:{0}, mismatches:{1}, indels:{2} ".format(matches, mismatches, indels))
# print(match_line)
result = aligner.align(x, y, revcomp=False)
y_alignment, match_line, x_alignment = result.alignment
# print()
# print(y_alignment)
# print(match_line)
# print(x_alignment)
matches, mismatches, indels = match_line.count("|"), match_line.count("*"), match_line.count(" ")
# alignment_length = len(match_line)
start_discrepancy = max(result.query_begin, result.reference_begin) # 0-indexed # max(result.query_begin, result.reference_begin) - min(result.query_begin, result.reference_begin)
query_end_discrepancy = len(x) - result.query_end - 1
ref_end_discrepancy = len(y) - result.reference_end - 1
end_discrepancy = max(query_end_discrepancy, ref_end_discrepancy) # max(result.query_end, result.reference_end) - min(result.query_end, result.reference_end)
# print("disc:", start_discrepancy, end_discrepancy)
tot_discrepancy = start_discrepancy + end_discrepancy
if 0 < start_discrepancy <= ends_discrepancy_threshold:
print("HERE",start_discrepancy)
matches_snippet = 0
mismatches_snippet = 0
if result.query_begin and result.reference_begin:
query_start_snippet = x[:result.query_begin]
ref_start_snippet = y[:result.reference_begin]
alns = pairwise2.align.globalds(query_start_snippet, ref_start_snippet, bio_matrix, g_open, g_extend)
top_aln = alns[0]
# print(alns)
mismatches_snippet = len(list(filter(lambda x: x[0] != x[1] and x[0] != '-' and x[1] != "-", zip(top_aln[0],top_aln[1]))))
indels_snippet = top_aln[0].count("-") + top_aln[1].count("-")
matches_snippet = len(top_aln[0]) - mismatches_snippet - indels_snippet
# print(matches_snippet, mismatches_snippet, indels_snippet)
query_start_alignment_snippet = top_aln[0]
ref_start_alignment_snippet = top_aln[1]
elif result.query_begin:
query_start_alignment_snippet = x[:result.query_begin]
ref_start_alignment_snippet = "-"*len(query_start_alignment_snippet)
indels_snippet = len(ref_start_alignment_snippet)
elif result.reference_begin:
ref_start_alignment_snippet = y[:result.reference_begin]
query_start_alignment_snippet = "-"*len(ref_start_alignment_snippet)
indels_snippet = len(query_start_alignment_snippet)
else:
print("BUG")
sys.exit()
matches, mismatches, indels = matches + matches_snippet, mismatches + mismatches_snippet, indels + indels_snippet
# print(ref_start_alignment_snippet)
# print(query_start_alignment_snippet)
y_alignment = ref_start_alignment_snippet + y_alignment
x_alignment = query_start_alignment_snippet + x_alignment
if 0 < end_discrepancy <= ends_discrepancy_threshold:
print("HERE2", end_discrepancy)
matches_snippet = 0
mismatches_snippet = 0
if query_end_discrepancy and ref_end_discrepancy:
query_end_snippet = x[result.query_end+1:]
ref_end_snippet = y[result.reference_end+1:]
alns = pairwise2.align.globalds(query_end_snippet, ref_end_snippet, bio_matrix, g_open, g_extend)
top_aln = alns[0]
mismatches_snippet = len(list(filter(lambda x: x[0] != x[1] and x[0] != '-' and x[1] != "-", zip(top_aln[0],top_aln[1]))))
indels_snippet = top_aln[0].count("-") + top_aln[1].count("-")
matches_snippet = len(top_aln[0]) - mismatches_snippet - indels_snippet
query_end_alignment_snippet = top_aln[0]
ref_end_alignment_snippet = top_aln[1]
elif query_end_discrepancy:
query_end_alignment_snippet = x[result.query_end+1:]
ref_end_alignment_snippet = "-"*len(query_end_alignment_snippet)
indels_snippet = len(ref_end_alignment_snippet)
elif ref_end_discrepancy:
ref_end_alignment_snippet = y[result.reference_end+1:]
query_end_alignment_snippet = "-"*len(ref_end_alignment_snippet)
indels_snippet = len(query_end_alignment_snippet)
else:
print("BUG")
sys.exit()
matches, mismatches, indels = matches + matches_snippet, mismatches + mismatches_snippet, indels + indels_snippet
y_alignment = y_alignment + ref_end_alignment_snippet
x_alignment = x_alignment + query_end_alignment_snippet
return x_alignment, y_alignment, matches, mismatches, indels, match_line
def create_isoform_graph(transcripts, min_exon):
G = nx.Graph()
for acc in transcripts.keys():
G.add_node(acc, accession = acc)
score_matrix = ssw.DNA_ScoreMatrix(match=1, mismatch=-2)
aligner = ssw.Aligner(gap_open=2, gap_extend=0, matrix=score_matrix)
cntr = 0
processed = set()
# already_assigned = set()
for acc1, seq1 in sorted(transcripts.items(), key= lambda x: len(x[1]), reverse=True):
cntr += 1
processed.add(acc1)
# print("length t:", len(seq1), acc1)
if cntr % 5 == 0:
print(cntr, "sequences processed")
# print(acc1)
# if acc1 in already_assigned:
# # print("allready assigned to larger sequence!")
# continue
for acc2, seq2 in sorted(transcripts.items(), key= lambda x: len(x[1]), reverse=True):
if acc2 in processed:
continue
# if acc1 == acc2:
# continue
# if seq1 == seq2:
# continue
# result = aligner.align(seq1, seq2, revcomp=False)
# seq2_aln, match_line, seq1_aln = result.alignment
# print(seq1)
# print(seq2)
seq1_aln, seq2_aln, matches, mismatches, indels, match_line = ssw_alignment(seq1, seq2, ends_discrepancy_threshold = 2000 )
print(acc1, acc2, mismatches, indels)
print(seq1_aln)
print(seq2_aln)
print(match_line)
# remove differences in 3' and 5' ends
tmp_seq1_aln = seq1_aln
tmp_seq1_aln = tmp_seq1_aln.lstrip("-")
tmp_seq1_aln = tmp_seq1_aln.rstrip("-")
tmp_seq2_aln = seq2_aln
tmp_seq2_aln = tmp_seq2_aln.lstrip("-")
tmp_seq2_aln = tmp_seq2_aln.rstrip("-")
del_seq1 = re.findall(r"[-]+",tmp_seq1_aln)
del_seq2 = re.findall(r"[-]+",tmp_seq2_aln)
mismatches = len([ 1 for n1, n2 in zip(seq1_aln,seq2_aln) if n1 != n2 and n1 != "-" and n2 != "-" ])
## do not count length discrepancies in ends
inner_del_seq1 = re.findall(r"[AGCT][-]+[AGCT]",seq1_aln)
inner_del_seq2 = re.findall(r"[AGCT][-]+[AGCT]",seq2_aln)
# print(inner_del_seq1)
# print(inner_del_seq2)
total_inner = sum([len(d) - 2 for d in inner_del_seq1]) + sum([len(d) - 2 for d in inner_del_seq2])
# print(indels, total_inner)
# by default (since all transcripts are distinct if we end up here), each transcript is its on gene member
# if we find an alingment that contains only structural changes of > X (2) nucleotides, and no other smaller differences we classify as same family
if mismatches == 0:
del_lengths1 = [len(del_) for del_ in del_seq1]
del_lengths2 = [len(del_) for del_ in del_seq2]
no_small_del_in_seq1 = ((len(del_lengths1) > 0 and min(del_lengths1) >= min_exon) or len(del_lengths1) == 0)
no_small_del_in_seq2 = ((len(del_lengths2) > 0 and min(del_lengths2) >= min_exon) or len(del_lengths2) == 0)
# print(no_small_del_in_seq1, no_small_del_in_seq2)
# print((len(del_seq1) > 0 and min(del_seq1) >= 3), len(del_seq1) == 0)
# if acc1[0][:14] == "transcript_460" and acc2[0][:14] == "transcript_467" :
# print("we are here", no_small_del_in_seq1, no_small_del_in_seq2, mismatches)
# sys.exit()
if no_small_del_in_seq1 and no_small_del_in_seq2:
G.add_edge(acc1, acc2, alignment={ acc1 : seq1_aln, acc2 : seq2_aln })
else:
pass
# print("Different only by small indel!!")
list_of_maximal_cliques = list(nx.find_cliques(G))
print("Number of possible members:", len(list_of_maximal_cliques) )
print("clique sizes", [ len(cl) for cl in sorted(list_of_maximal_cliques, key= lambda x: len(x), reverse=True)] )
return G