def get_pileup_list(all_sample_read_aligns, sample_names, reference_graph, node_grouping_list=None, show_insertions=False):
""" Generate genotype_pileups for all sample
Args:
all_sample_read_aligns: dict: {sample_name : {repeat_id: [list of read aligns]}}
sample_names: ordered list of sample names in pileup
reference_graph: ReferenceGraph for STR locus
node_grouping_list: Ordered list of ReferenceNodes in decreasing order of priority. (Default: left to right, repeat units first)
show_insertions: Boolean to show full sequence of insertions
Returns:
List of 3-tuples with pileups for each sample
e.g. [(sample_name, PathWithInsertions, Pileup (see get_pileup))]
"""
for s in all_sample_read_aligns:
repeat_aligns = [a for a in all_sample_read_aligns[s] if len([n for n in parse_graph_cigar(a.graph_cigar) if reference_graph[int(n[0])].is_repeat]) > 0]
all_sample_read_aligns[s] = repeat_aligns
repeat_path = get_repeat_path(all_sample_read_aligns, reference_graph)
pileup_list = []
for sample_name in sample_names:
if sample_name in all_sample_read_aligns:
genotype_pileup_list = get_pileup(all_sample_read_aligns[sample_name], reference_graph, repeat_path, node_grouping_list, show_insertions=show_insertions)
else:
genotype_pileup_list = get_pileup([], reference_graph, repeat_path, node_grouping_list, show_insertions=show_insertions)
pileup_list.append((sample_name, repeat_path, genotype_pileup_list[1]))
return pileup_list
def get_spanning_genotypes(read_align_list, reference_graph):
'''From a list of ReadAlignments, return a dict mapping each repetitive node_id to ordered list of genotypes spanned by a read.
Arguments:
read_align_list: List of ReadAlignment to the locus in the sample
reference_graph: ReferenceGraph for STR locus
node_grouping_list: Ordered list of ReferenceNodes in decreasing order of priority. (Default: left to right, repeat units first)
Returns:
dict mapping node_id to list of spanned genotypes
'''
spanning_genotypes = defaultdict(lambda: [], {})
for read_alignment in read_align_list:
node_count = np.zeros(len(reference_graph))
left_flank_uncovered = np.ones(len(reference_graph))
right_flank_uncovered = np.ones(len(reference_graph))
previous_node_id = -1
for node in parse_graph_cigar(read_alignment.graph_cigar):
node_count[int(node[0])] += 1
if previous_node_id != -1 and previous_node_id != int(node[0]):
for i in range(previous_node_id, int(node[0])):
right_flank_uncovered[i] = 0
left_flank_uncovered[i + 1] = 0
previous_node_id = int(node[0])
for node in reference_graph:
if node.is_repeat and right_flank_uncovered[node.node_id] == 0 and left_flank_uncovered[node.node_id] == 0:
if node_count[node.node_id] not in spanning_genotypes[node.node_id]:
spanning_genotypes[node.node_id].append(node_count[node.node_id])
for node_id in spanning_genotypes:
spanning_genotypes[node_id].sort()
return spanning_genotypes
def get_repeat_path(read_alignments, reference_graph):
"""Get a PathWithInsertions corresponding to highest count of each repeat_unit
Arguments:
read_alignments: dict mapping sample name to list of ReadAlignments
reference_graph: ReferenceGraph for repeat locus
"""
#TODO: Infer repeat_graph from read_alignments if repeat_specs not provided.
max_node_count = defaultdict(lambda: 0, {})
for sample in read_alignments:
for read_alignment in read_alignments[sample]:
node_count = defaultdict(lambda: 0, {})
for node in parse_graph_cigar(read_alignment.graph_cigar):
node_count[int(node[0])] += 1
for node_id in node_count:
max_node_count[node_id] = max(max_node_count[node_id], node_count[node_id])
if max_node_count[reference_graph[0].node_id] == 0:
max_node_count[reference_graph[0].node_id] = 1
if max_node_count[reference_graph[-1].node_id] == 0:
max_node_count[reference_graph[-1].node_id] = 1
node_id_list = []
for node in reference_graph:
node_id_list += [node.node_id] * max_node_count[node.node_id]
repeat_path = PathWithInsertions(node_id_list)
print ("Reference path = %s" % [n.node_id for n in repeat_path])
for sample in read_alignments:
for read_alignment in read_alignments[sample]:
repeat_path.update(read_alignment.graph_cigar, read_alignment.offset)
return repeat_path
def get_alignment_coordinates(self,
query_sequence,
reference_graph,
graph_cigar,
offset=0,
show_insertions=False):
"""Update nodes with insertions in the read alignment.
Arguments:
query_sequence: sequence of aligned read
reference_graph: ReferenceGraph for locus
graph_cigar: graph cigar string of read alignment.
offset: 0-based position of first base in the first node in the alignment
show_insertions: Boolean to show full sequence of insertions
Returns:
alignment_coordinates: a list of 3-tuples
Each 3-tuple consists of (coordinate, query_bp, ref_bp)
Gaps are represented by "-"
"""
INTERNODE_GAP = 1
alignment_coordinates = []
graph_cigar_tuples = parse_graph_cigar(graph_cigar)
path_index = self.get_alignment_start_index(graph_cigar)
node_coordinate = 0
cigar_tuple_index = 0
query_pos = 0
for i in range(path_index):
node_coordinate += len(reference_graph[self[i].node_id].seq)
if show_insertions:
node_coordinate += self[i].get_total_insertion_size()
node_coordinate += INTERNODE_GAP
for node_alignment in graph_cigar_tuples:
while path_index < len(self) and self[path_index].node_id != int(
node_alignment[0]):
node_coordinate += len(
reference_graph[self[path_index].node_id].seq)
if show_insertions:
node_coordinate += self[
path_index].get_total_insertion_size()
path_index += 1
node_coordinate += INTERNODE_GAP
cigar = node_alignment[1]
#TODO: can we rely on graph cigar to always include bases from node extremes? Then we do not need offset
if cigar_tuple_index == 0:
node_offset = offset
else:
node_offset = 0
node_align = self[path_index].get_alignment_coordinates(
cigar, query_pos, query_sequence, node_coordinate, node_offset,
reference_graph, show_insertions)
alignment_coordinates += node_align
query_pos += get_cigar_query_align_len(cigar)
cigar_tuple_index += 1
node_coordinate += len(
reference_graph[self[path_index].node_id].seq)
if show_insertions:
node_coordinate += self[path_index].get_total_insertion_size()
node_coordinate += INTERNODE_GAP
path_index += 1
return alignment_coordinates
def get_genotype_classification(read_alignment, reference_graph, spanning_genotypes, node_grouping_list=None):
'''From a list of ReadAlignments, return a dict mapping each repetitive node_id to ordered list of genotypes spanned by a read.
Arguments:
read_align_list: List of ReadAlignment to the locus in the sample
reference_graph: ReferenceGraph for STR locus
node_grouping_list: Ordered list of ReferenceNodes in decreasing order of priority. (Default: left to right, repeat units first)
Returns:
list of 3 tuples for each node in node_grouping_list: (ReferenceNode, CompareStatus ("=", ">" or "\leq"), Genotype String)
'''
node_count = np.zeros(len(reference_graph))
left_flank_uncovered = np.ones(len(reference_graph))
right_flank_uncovered = np.ones(len(reference_graph))
previous_node_id = -1
for cigar_node in parse_graph_cigar(read_alignment.graph_cigar):
node_count[int(cigar_node[0])] += 1
if previous_node_id != -1 and previous_node_id != int(cigar_node[0]):
for i in range(previous_node_id, int(cigar_node[0])):
right_flank_uncovered[i] = 0
left_flank_uncovered[i + 1] = 0
previous_node_id = int(cigar_node[0])
gt_class = []
if node_grouping_list is None:
node_grouping_list = ([n.node_id for n in reference_graph if (n.is_repeat and 'ignore' not in n.node_name)]
+ [n.node_id for n in reference_graph if (n.is_repeat and 'ignore' in n.node_name)]
+ [n.node_id for n in reference_graph if not n.is_repeat])
for node_id in node_grouping_list:
graph_node = reference_graph[node_id]
if not graph_node.is_repeat:
continue
gt_compare_status = "="
if (node_count[graph_node.node_id], left_flank_uncovered[graph_node.node_id], right_flank_uncovered[graph_node.node_id]) == (0, 1, 1):
gt = -1
elif ((left_flank_uncovered[graph_node.node_id] != right_flank_uncovered[graph_node.node_id]) or
node_count[graph_node.node_id] > 0 and (left_flank_uncovered[graph_node.node_id], right_flank_uncovered[graph_node.node_id]) == (1, 1)):
spanning_index = bisect.bisect_left(spanning_genotypes[graph_node.node_id], node_count[graph_node.node_id])
if spanning_index >= len(spanning_genotypes[graph_node.node_id]):
gt_compare_status = '>'
if len(spanning_genotypes[graph_node.node_id]) == 0:
gt = 0
else:
gt = spanning_genotypes[graph_node.node_id][-1]
else:
gt = spanning_genotypes[graph_node.node_id][spanning_index]
gt_compare_status = '\leq'
else:
gt = node_count[graph_node.node_id]
gt_string = r'$gt%s %s %s$' % (graph_node.node_id, gt_compare_status, int(gt))
if gt > 0 and gt_compare_status == '\leq':
spanning_index = bisect.bisect_left(spanning_genotypes[graph_node.node_id], node_count[graph_node.node_id])
if spanning_index > 0:
gt_string = r"$%s < gt%s \leq %s$" % (int(spanning_genotypes[graph_node.node_id][spanning_index - 1]), graph_node.node_id, int(gt))
gt_class.append((graph_node, gt_compare_status, gt, gt_string))
return gt_class
def update(self, graph_cigar, offset=0):
"""Update nodes with insertions in the read alignment.
Arguments:
graph_cigar: graph cigar string of read alignment.
offset: 0-based position of first base in the first node in the alignment
"""
graph_cigar_tuples = parse_graph_cigar(graph_cigar)
path_index = self.get_alignment_start_index(graph_cigar)
cigar_tuple_index = 0
for node_alignment in graph_cigar_tuples:
while path_index < len(self) and self[path_index].node_id != int(
node_alignment[0]):
path_index += 1
if cigar_tuple_index == 0:
self[path_index].update(node_alignment[1], offset)
else:
self[path_index].update(node_alignment[1])
cigar_tuple_index += 1
path_index += 1
def get_graph_cigar_seq(self, graph_cigar, offset=0):
'''
Get reference sequence corresponding to graph_cigar alignment
Args:
graph_cigar: graph_cigar string
offset: first matched position in the first node (0-based)
Returns:
reference sequence
'''
graph_cigar_seq = ''
graph_cigar_tuples = seq_util.parse_graph_cigar(graph_cigar)
for graph_cigar_tuple in enumerate(graph_cigar_tuples):
if graph_cigar_tuple[0] == 0:
graph_cigar_seq += self[int(
graph_cigar_tuple[1][0])].get_cigar_seq(
graph_cigar_tuple[1][1], offset)
else:
graph_cigar_seq += self[int(
graph_cigar_tuple[1][0])].get_cigar_seq(
graph_cigar_tuple[1][1])
return graph_cigar_seq
def get_alignment_start_index(self, graph_cigar):
"""Find index of first node in alignment from graph_cigar string
Arguments:
graph_cigar: graph cigar string
Returns:
Ordered list of node indices in PathWithInsertions.
"""
graph_cigar_tuples = parse_graph_cigar(graph_cigar)
if len(graph_cigar_tuples) == 0:
return 0
first_id = int(graph_cigar_tuples[0][0])
# left aligned if single node or in-repeat
if first_id == int(graph_cigar_tuples[-1][0]):
first_start_index = 0
while first_start_index < len(self):
current_node = self[first_start_index]
if current_node.node_id == first_id:
return first_start_index
first_start_index += 1
cigar_index = 0
while cigar_index < len(graph_cigar_tuples):
if int(graph_cigar_tuples[cigar_index][0]) != int(
graph_cigar_tuples[0][0]):
break
cigar_index += 1
first_count = cigar_index
first_end_index = 0
current_node_id = self[first_end_index].node_id
while first_end_index < len(self):
if first_end_index == len(self) - 1:
return first_end_index - first_count + 1
next_node_id = self[first_end_index + 1].node_id
if current_node_id == first_id and next_node_id != first_id:
return first_end_index - first_count + 1
first_end_index += 1
current_node_id = next_node_id
alignment_start_index = first_end_index - first_count + 1
return alignment_start_index