def get_edges_from_gfa(gfa_fpath, output_dirpath, min_edge_len):
if not gfa_fpath:
return None
input_edges_fpath = join(dirname(gfa_fpath), get_filename(gfa_fpath) + ".fasta")
edges_fpath = join(output_dirpath, basename(input_edges_fpath))
if not is_empty_file(gfa_fpath) and not can_reuse(edges_fpath, files_to_check=[gfa_fpath]):
print("Extracting edge sequences from " + gfa_fpath + "...")
with open(edges_fpath, "w") as out:
with open(gfa_fpath) as f:
for line in f:
if line.startswith('S'):
fs = line.strip().split()
seq_name = fs[1]
seq = None
if is_acgt_seq(fs[2]):
seq = fs[2]
elif len(fs) >= 4 and is_acgt_seq(fs[3]):
seq = fs[3]
if seq and len(seq) >= min_edge_len:
out.write(">%s\n" % get_edge_agv_id(get_edge_num(seq_name)))
out.write(seq)
out.write("\n")
if is_empty_file(edges_fpath) and not is_empty_file(input_edges_fpath):
with open(edges_fpath, "w") as out:
with open(input_edges_fpath) as f:
for line in f:
if line.startswith('>'):
seq_name = line.strip().split()[0][1:]
out.write(">%s\n" % get_edge_agv_id(get_edge_num(seq_name)))
else:
out.write(line)
return edges_fpath
def parse_canu_unitigs_info(input_dirpath, dict_edges):
tiginfo_fpath = find_file_by_pattern(input_dirpath, ".unitigs.layout.tigInfo")
if not is_empty_file(tiginfo_fpath):
with open(tiginfo_fpath) as f:
for i, line in enumerate(f):
if i == 0:
header = line.strip().split()
repeat_col = header.index("sugRept") if "sugRept" in header else None
cov_col = header.index("coverage") if "coverage" in header else None
if repeat_col is None or cov_col is None:
break
continue
fs = line.strip().split()
edge_id = get_edge_agv_id(get_edge_num(fs[0]))
rc_edge_id = get_edge_agv_id(-get_edge_num(fs[0]))
if edge_id in dict_edges:
coverage = int(float(fs[cov_col]))
dict_edges[edge_id].cov = coverage
dict_edges[rc_edge_id].cov = coverage
if fs[repeat_col] == "yes":
dict_edges[edge_id].repetitive = True
dict_edges[rc_edge_id].repetitive = True
# else:
# print("Warning! Edge %s is not found!" % edge_id)
return dict_edges
def parse_canu_contigs_info(input_dirpath):
contig_info = dict()
edges_by_contig = defaultdict(list)
unitigs_info_fpath = find_file_by_pattern(input_dirpath, "unitigs.bed")
if input_dirpath and not is_empty_file(unitigs_info_fpath):
with open(unitigs_info_fpath) as f:
for line in f:
fs = line.strip().split()
contig, start, end, unitig = fs[:4]
strand = fs[-1]
edge_name = get_edge_num(
unitig) if strand == "+" else -get_edge_num(unitig)
contig_id = get_canu_id(contig)
edges_by_contig[contig_id].append(str(edge_name))
contigs_info_fpath = find_file_by_pattern(input_dirpath,
"contigs.layout.tigInfo")
if input_dirpath and not is_empty_file(contigs_info_fpath):
len_col = None
cov_col = None
with open(contigs_info_fpath) as f:
for i, line in enumerate(f):
if i == 0:
header = line.strip().split()
len_col = header.index(
"tigLen") if "tigLen" in header else None
cov_col = header.index(
"coverage") if "coverage" in header else None
if len_col is None or cov_col is None:
break
continue
fs = line.strip().split()
length = int(float(fs[len_col]))
coverage = int(float(fs[cov_col]))
contig_id = get_canu_id(fs[0])
if contig_id in edges_by_contig:
contig_info[contig_id] = {
'length': length,
'cov': coverage,
'mult': 1
}
for contig_id, edges in edges_by_contig.items():
contig_info[contig_id]['edges'] = edges
return contig_info
def format_edges_file(input_fpath, output_dirpath):
if is_empty_file(input_fpath):
return None
edges_fpath = join(output_dirpath, "edges.fasta")
if not can_reuse(edges_fpath, files_to_check=[input_fpath]):
with open(input_fpath) as f:
with open(edges_fpath, "w") as out_f:
for line in f:
if line.startswith('>'):
edge_id = get_edge_agv_id(get_edge_num(line[1:]))
out_f.write(">%s\n" % edge_id)
else:
out_f.write(line)
return edges_fpath
def parse_canu_assembly_info(input_dirpath, dict_edges):
contig_edges = defaultdict(list)
unitigs_fpath = find_file_by_pattern(input_dirpath, ".unitigs.bed")
if is_empty_file(unitigs_fpath):
print(
"Warning! Unitigs.bed is not found, information about contigs will not be provided"
)
with open(unitigs_fpath) as f:
for line in f:
fs = line.strip().split()
contig, start, end, unitig = fs[:4]
edge_id = get_edge_agv_id(get_edge_num(unitig))
if edge_id in dict_edges:
contig_id = get_canu_id(contig)
contig_edges[contig_id].append((start, end, edge_id))
return contig_edges
def parse_gfa(gfa_fpath, min_edge_len, input_dirpath=None, assembler=None):
dict_edges = dict()
predecessors = defaultdict(list)
successors = defaultdict(list)
g = nx.DiGraph()
print("Parsing " + gfa_fpath + "...")
# gfa = gfapy.Gfa.from_file(gfa_fpath, vlevel = 0)
links = []
edge_overlaps = defaultdict(dict)
with open(gfa_fpath) as f:
for line in f:
record_type = line[0]
if record_type == 'S':
fs = line.split()
name, seq_len = fs[1], len(fs[2])
if fs[2] == '*':
seq_len = None
add_fields = fs[3:] if len(fs) > 3 else []
add_info = dict((f.split(':')[0].lower(), f.split(':')[-1]) for f in add_fields)
cov = 1
if "dp" in add_info:
cov = float(add_info["dp"]) ## coverage depth
elif "kc" in add_info:
cov = max(1, int(add_info["kc"]) / seq_len) ## k-mer count / edge length
if "ln" in add_info:
seq_len = int(add_info["ln"]) ## sequence length
if seq_len and seq_len >= min_edge_len:
edge_id = get_edge_agv_id(get_edge_num(name))
edge = Edge(edge_id, get_edge_num(name), seq_len, cov, element_id=edge_id)
dict_edges[edge_id] = edge
for overlapped_edge, overlap in edge_overlaps[edge_id].items():
dict_edges[edge_id].overlaps.append((edge_id_to_name(overlapped_edge), overlapped_edge, overlap))
rc_edge_id = get_edge_agv_id(-get_edge_num(name))
rc_edge = Edge(rc_edge_id, -get_edge_num(name), seq_len, cov, element_id=rc_edge_id)
dict_edges[rc_edge_id] = rc_edge
for overlapped_edge, overlap in edge_overlaps[rc_edge_id].items():
dict_edges[edge_id].overlaps.append((edge_id_to_name(overlapped_edge), overlapped_edge, overlap))
if record_type != 'L' and record_type != 'E':
continue
if record_type == 'L':
_, from_name, from_orient, to_name, to_orient = line.split()[:5]
else:
# E * 2+ 65397+ 21 68$ 0 47 47M
from_name, to_name = line.split()[2], line.split()[3]
from_orient, to_orient = from_name[-1], to_name[-1]
from_name, to_name = from_name[:-1], to_name[:-1]
edge1 = get_edge_agv_id(get_edge_num(from_name))
edge2 = get_edge_agv_id(get_edge_num(to_name))
if from_orient == '-': edge1 = get_match_edge_id(edge1)
if to_orient == '-': edge2 = get_match_edge_id(edge2)
overlap = 0
overlap_operations = re.split('(\d+)', line.split()[-1].strip())
for i in range(0, len(overlap_operations) - 1, 1):
if not overlap_operations[i]:
continue
if overlap_operations[i+1] == 'M' or overlap_operations[i+1] == 'I':
overlap += int(overlap_operations[i])
links.append((from_name, from_orient, to_name, to_orient, overlap))
if overlap:
edge_overlaps[edge1][edge2] = overlap
edge_overlaps[edge2][edge1] = overlap
### gfa retains only canonical links
for link in links:
from_name, from_orient, to_name, to_orient, overlap = link
edge1 = get_edge_agv_id(get_edge_num(from_name))
edge2 = get_edge_agv_id(get_edge_num(to_name))
if from_orient == '-': edge1 = get_match_edge_id(edge1)
if to_orient == '-': edge2 = get_match_edge_id(edge2)
if edge1 != edge2:
predecessors[edge2].append(edge1)
successors[edge1].append(edge2)
g.add_edge(edge1, edge2)
if is_spades(assembler) or is_abyss(assembler):
edge1, edge2 = get_match_edge_id(edge2), get_match_edge_id(edge1)
if edge1 != edge2:
predecessors[edge2].append(edge1)
successors[edge1].append(edge2)
g.add_edge(edge1, edge2)
if assembler == "canu" and input_dirpath:
dict_edges = parse_canu_unitigs_info(input_dirpath, dict_edges)
dict_edges = construct_graph(dict_edges, predecessors, successors)
print("Finish parsing.")
return dict_edges
def run_quast_analysis(input_fpath,
reference_fpath,
output_dirpath,
json_output_dirpath,
threads,
contig_edges,
dict_edges=None,
is_meta=False):
ms_out_fpath = None
quast_output_dir = join(
output_dirpath,
"quast_output" if not dict_edges else "quast_edge_output")
if not is_empty_file(input_fpath) and not is_empty_file(reference_fpath):
ms_out_fpath = get_mis_report_fpath(quast_output_dir, input_fpath)
ms_out_fpath = run(input_fpath, reference_fpath, ms_out_fpath,
quast_output_dir, threads, is_meta)
if not ms_out_fpath:
if not is_empty_file(input_fpath) and not is_empty_file(
reference_fpath):
print(
"QUAST failed! Make sure you are using the latest version of QUAST"
)
print("No information about %s mappings to the reference genome" %
("edge" if dict_edges else "contig"))
with open(join(json_output_dirpath, "reference.json"), 'w') as handle:
handle.write("chrom_lengths=" + json.dumps([]) + ";\n")
handle.write("edgeMappingInfo=" + json.dumps([]) + ";\n")
handle.write("chromGaps=" + json.dumps([]) + ";\n")
handle.write("chromAligns=" + json.dumps([]) + ";\n")
with open(join(json_output_dirpath, 'errors.json'), 'w') as handle:
handle.write("misassembledContigs=[];\n")
return None, None, None, dict_edges
# search for misassemblies and store them for each edge and contig
misassembled_seqs = defaultdict(list)
with open(ms_out_fpath) as f:
seq_id = ''
for line in f:
if line.startswith("Extensive misassembly"):
match = re.search(align_pattern, line)
if not match or len(match.groups()) < 4:
continue
start1, end1, start2, end2 = match.group(
'start1'), match.group('end1'), match.group(
'start2'), match.group('end2')
if dict_edges:
edge_id = get_edge_agv_id(get_edge_num(seq_id))
dict_edges[edge_id].errors.append(
(start1, end1, start2, end2))
else:
misassembled_seqs[seq_id].append(
(start1, end1, start2, end2))
## add misassembl edge
else:
seq_id = line.strip()
if not dict_edges:
with open(join(json_output_dirpath, 'errors.json'), 'w') as handle:
handle.write("misassembledContigs='" +
json.dumps(misassembled_seqs) + "';\n")
return None, None, None, dict_edges
else:
parse_alignments(get_alignments_fpath(quast_output_dir, input_fpath),
json_output_dirpath)
mapping_fpath = map_edges_to_ref(input_fpath, output_dirpath,
reference_fpath, threads)
mapping_info, chrom_names, edge_by_chrom = parse_mapping_info(
mapping_fpath, json_output_dirpath, dict_edges)
return mapping_info, chrom_names, edge_by_chrom, dict_edges
def parse_mapping_info(mapping_fpath, json_output_dir, dict_edges):
# assign edges to chromosomes and color edges to corresponding colors
mapping_info = defaultdict(set)
edge_mappings = defaultdict(lambda: defaultdict(list))
edge_lengths = dict()
chrom_lengths = dict()
with open(mapping_fpath) as f:
for line in f:
# contig_1 257261 14 160143 - chr13 924431 196490 356991 147365 161095 60 tp:A:P cm:i:14049 s1:i:147260 s2:i:4375 dv:f:0.0066
fs = line.split()
edge_id = get_edge_agv_id(get_edge_num(fs[0]))
start, end = int(fs[2]), int(fs[3])
edge_lengths[edge_id] = int(fs[1])
chrom, chrom_len = fs[5], int(fs[6])
ref_start, ref_end = int(fs[7]), int(fs[8])
chrom_lengths[chrom] = chrom_len
edge_mappings[edge_id][chrom].append(
(start, end, ref_start, ref_end))
chroms_by_edge = defaultdict(set)
edge_by_chrom = defaultdict(set)
chrom_names = set()
best_aligns = defaultdict(defaultdict)
for edge_id in edge_mappings:
# assign an edge to a chromosome if more than 90% of edge aligned to the chromosome
len_threshold = 0.9 * edge_lengths[edge_id]
gap_threshold = min(5000, 0.05 * edge_lengths[edge_id])
for chrom, mappings in edge_mappings[edge_id].items():
mappings.sort(key=lambda x: (x[0], -x[1]), reverse=False)
for chrom, mappings in edge_mappings[edge_id].items():
aligns = []
covered_len = 0
last_pos = 0
last_ref_pos = 0
align_s, align_e = 0, 0
# calculate covered length (do not count overlaps)
for (start, end, ref_start, ref_end) in mappings:
start = max(start, last_pos)
covered_len += max(0, end - start + 1)
last_pos = max(last_pos, end + 1)
if covered_len >= len_threshold:
chroms_by_edge[edge_id].add(chrom)
chrom_names.add(chrom)
edge_by_chrom[chrom].add(edge_id)
mappings.sort(key=lambda x: (x[2], -x[3]), reverse=False)
for (start, end, ref_start, ref_end) in mappings:
ref_start = max(ref_start, last_ref_pos)
last_ref_pos = max(last_ref_pos, ref_end + 1)
if not align_s:
align_s = ref_start
if align_e and ref_start - align_e >= gap_threshold:
if align_e - align_s >= 500: # break alignments if gap longer than 500 bp
aligns.append((chrom, align_s, align_e))
align_s = ref_start
align_e = ref_end - 1
if align_e and align_e - align_s >= 500:
aligns.append((chrom, align_s, align_e))
aligns.sort(reverse=True, key=lambda x: x[2] - x[1])
edge_alignment = chrom + ":"
if aligns:
best_aligns[edge_id][chrom] = aligns[0][1]
for align in aligns[:
3]: # store top 3 alignments for each edge
edge_alignment += " %s-%s," % (format_pos(
align[1]), format_pos(align[2]))
dict_edges[edge_id].aligns[chrom] = edge_alignment[:-1]
if get_match_edge_id(edge_id) in dict_edges:
dict_edges[get_match_edge_id(
edge_id)].aligns[chrom] = edge_alignment[:-1]
chrom_len_dict = OrderedDict(
(chrom, chrom_lengths[chrom])
for i, chrom in enumerate(list(natural_sort(chrom_names))))
non_alt_chroms = [
c for c in chrom_names
if 'alt' not in c and 'random' not in c and 'chrUn' not in c
]
chrom_order = OrderedDict(
(chrom, i)
for i, chrom in enumerate(list(natural_sort(non_alt_chroms))))
color_list = [
'#e6194b', '#3cb44b', '#ffe119', '#1792d4', '#f58231', '#911eb4',
'#46f0f0', '#f032e6', '#d2f53c', '#fabebe', '#00dbb1', '#dba2ff',
'#aa6e28', '#83360e', '#800000', '#003bff', '#808000', '#8d73d4',
'#000080', '#806680', '#51205a', '#558859', '#d1a187', '#87a1d1',
'#87a1d1', '#afd187'
]
edge_chroms = defaultdict(set)
for edge_id, chroms in chroms_by_edge.items():
match_edge_id = edge_id.replace(
'rc', 'e') if edge_id.startswith('rc') else edge_id.replace(
'e', 'rc')
for chrom in chroms:
edge_chroms[edge_id].add(chrom)
edge_chroms[match_edge_id].add(chrom)
if match_edge_id in dict_edges:
edge_by_chrom[chrom].add(match_edge_id)
is_single_chrom = len(chrom_order.keys()) == 1
for edge_id, chroms in edge_chroms.items():
if edge_id not in dict_edges:
continue
mapping_info[edge_id] = list(chroms)
colors = set()
for chrom in chroms:
if chrom in chrom_order:
if is_single_chrom: # color an edge according to its position in reference
pos = best_aligns[edge_id][chrom] if best_aligns[edge_id] else \
best_aligns[get_match_edge_id(edge_id)][chrom]
color = get_rainbow_color(pos, chrom_len_dict[chrom])
else:
color = color_list[chrom_order[chrom] % len(color_list)]
else:
color = '#808080'
colors.add(color)
if len(colors) <= 5:
dict_edges[edge_id].chrom = ':'.join(list(colors))
else:
dict_edges[edge_id].chrom = 'white:red:black:red:black:white'
with open(join(json_output_dir, "reference.json"), 'a') as handle:
handle.write("chrom_lengths=" + json.dumps(chrom_len_dict) + ";\n")
handle.write("edgeMappingInfo=" + json.dumps(mapping_info) + ";\n")
return mapping_info, non_alt_chroms, edge_by_chrom