def combine_nodes(graph: GraphContainer):
"""
Combine adjacent nodes with the same sequence labels
"""
for n1 in list(graph.nodes.values()):
if len(list(graph.outEdges(n1))) != 1:
continue # Pair of nodes with no other in/out edges
n2 = graph.nodes[next(graph.outEdges(n1))["to"]]
if len(list(graph.inEdges(n2))) != 1:
continue
if not (n1["chrom"] == n2["chrom"] and n1["end"] + 1 == n2["start"]):
continue # nodes must be adjacent
haplos = n1["sequences"]
if n2["sequences"] != haplos:
continue # only collapse nodes with same haplotypes
if "reference" in n1:
if "reference" not in n2:
continue # nodes must be of same type
node = graph.add_refNode(n1["chrom"], n1["start"], n2["end"], haplos)
else:
if "reference" in n2:
continue # nodes must be of same type
node = graph.add_altNode(
n1["chrom"], n1["start"], n2["end"], n1["sequence"] + n2["sequence"], haplos)
logging.info("Combinding %s and %s", n1['name'], n2['name'])
for e in list(graph.inEdges(n1)):
graph.add_edge(graph.nodes[e["from"]], node, e["sequences"])
for e in list(graph.outEdges(n2)):
graph.add_edge(node, graph.nodes[e["to"]], e["sequences"])
graph.del_node(n1)
graph.del_node(n2)
def add_source_sink(graph: GraphContainer,
source_name="source",
sink_name="sink"):
"""
add source and sink if necessary and link to exising nodes without in/out edges
:param graph: graph to work on
:param source_name: name of source node
:param sink_name: name of sink node
"""
if source_name not in graph.nodes:
graph.nodes[source_name] = {
"name": source_name,
"sequence": "N" * 10
}
if sink_name not in graph.nodes:
graph.nodes[sink_name] = {
"name": sink_name,
"sequence": "N" * 10
}
# Link nodes without incoming/outgoing edges to source/sink
for node in graph.nodes.values():
if node["name"] in [source_name, sink_name]:
continue
if not any(graph.inEdges(node)):
logging.info("Linking %s from source", node['name'])
graph.add_edge(graph.nodes[source_name], node)
if not any(graph.outEdges(node)):
logging.info("Linking %s to sink", node['name'])
graph.add_edge(node, graph.nodes[sink_name])
def split_alt_nodes(graph: GraphContainer, max_len=300, padding_len=150):
"""
Split long alternate nodes
:param graph: graph to work on
:param max_len: max length of reference node with no sequences
:param padding_len: length of sequence to keep
"""
assert max_len >= 2 * padding_len
for node in list(graph.altNodes()):
if len(node["sequence"]) <= max_len:
continue
logging.info(f"Splitting long ALT node: {node['name']}")
n1 = graph.add_altNode(node["chrom"], node["start"], node["end"],
node["sequence"][:padding_len],
node["sequences"])
n2 = graph.add_altNode(node["chrom"], node["start"], node["end"],
node["sequence"][-padding_len:],
node["sequences"])
for e in list(graph.inEdges(node)):
graph.add_edge(graph.nodes[e["from"]], n1, e["sequences"])
for e in list(graph.outEdges(node)):
graph.add_edge(n2, graph.nodes[e["to"]], e["sequences"])
graph.del_node(node)
def remove_empty_nodes(graph: GraphContainer):
"""
Remove nodes without sequence (from deletions / skipped insertions or split ref nodes)
Merge in & out edge pairs to keep connections
"""
for node in list(graph.nodes.values()):
if (("reference" in node and node["start"] <= node["end"]) or
node.get("sequence", "") != ""):
continue
logging.info("Removing empty node %s", node['name'])
inSeqs = [s for e in graph.inEdges(node) for s in e["sequences"]]
outSeqs = [s for e in graph.outEdges(node) for s in e["sequences"]]
for e1 in list(graph.inEdges(node)):
for e2 in list(graph.outEdges(node)):
# Label the new edges with sequence labels either observed
# on both merged in- and out-edge or on an in (out) -edge only
# if the label is undetermined goung out (in)
haplos = e1["sequences"].intersection(e2["sequences"]).union(
e1["sequences"].difference(outSeqs).union(
e2["sequences"].difference(inSeqs)))
graph.add_edge(graph.nodes[e1["from"]], graph.nodes[e2["to"]], haplos)
graph.del_node(node)
def split_node(graph: GraphContainer, node, breakpoints):
"""
Split a node at a set of breakpoints and link new (sub-)nodes
Used to link to new variant nodes later
Modifies graph and deletes node after splitting
:returns Created sub-nodes
"""
if not breakpoints:
return node
breakpoints = sorted(set(breakpoints))
logging.debug(f"Splitting {node['name']} at {breakpoints}")
nodes = []
lEnd = 0
for p in breakpoints:
assert 0 <= p <= node["end"] - node["start"] + 1
nStart = node["start"] + lEnd
nEnd = node["start"] + p - 1
if "reference" in node:
nodes.append(
graph.add_refNode(node["chrom"], nStart, nEnd,
node["sequences"]))
else:
seq = node["sequence"][lEnd:p]
nodes.append(
graph.add_altNode(node["chrom"], nStart, nEnd, seq,
node["sequences"]))
lEnd = p
# Add last node
lStart = node["start"] + breakpoints[-1]
if "reference" in node:
nodes.append(
graph.add_refNode(node["chrom"], lStart, node["end"],
node["sequences"]))
else:
seq = node["sequence"][breakpoints[-1]:]
nodes.append(
graph.add_altNode(node["chrom"], lStart, node["end"], seq,
node["sequences"]))
# Connect nodes
for e in graph.inEdges(node):
graph.add_edge(graph.nodes[e["from"]], nodes[0], e["sequences"])
for e in graph.outEdges(node):
graph.add_edge(nodes[-1], graph.nodes[e["to"]], e["sequences"])
for (n1, n2) in zip(nodes[:-1], nodes[1:]):
graph.add_edge(n1, n2)
# Delete original node, unless identical to new node (no split)
if node['name'] not in [n['name'] for n in nodes]:
graph.del_node(node)
return nodes
def split_ref_nodes(graph: GraphContainer, max_len=300, padding_len=150):
"""
Split long reference nodes
:param graph: graph to work on
:param max_len: max length of reference node with no sequences
:param padding_len: length of sequence to keep
"""
assert max_len >= 2 * padding_len
for node in list(graph.refNodes()):
if node["end"] - node["start"] + 1 <= max_len:
continue
logging.info("Splitting long REF node: %s", node['name'])
firstEnd = node["start"] + padding_len - 1
n1 = graph.add_refNode(node["chrom"], node["start"], firstEnd, node["sequences"])
sndStart = node["end"] - padding_len + 1
n2 = graph.add_refNode(node["chrom"], sndStart, node["end"], node["sequences"])
for e in list(graph.inEdges(node)):
graph.add_edge(graph.nodes[e["from"]], n1, e["sequences"])
for e in list(graph.outEdges(node)):
graph.add_edge(n2, graph.nodes[e["to"]], e["sequences"])
graph.del_node(node)
def get_graph(self, allele_graph=False):
""" Create the paragraph representation of nodes and edges for this graph
:param alleleGraph: create edges between any compatible allele pair (rather
than just following reference and given haplotypes)
:return GraphContainer object
"""
logging.info("Creating output graph")
graph = GraphContainer()
# create ref nodes
pnode = None
for ref in self.get_ref_alleles():
node = graph.add_refNode(self.chrom, ref.begin, ref.end - 1,
ref.data.haplotypes)
if pnode:
assert pnode["end"] + 1 == node["start"]
graph.add_edge(pnode, node)
pnode = node
# Create alt nodes
for alt in self.alts.values():
graph.add_altNode(self.chrom, alt.start, alt.end, alt.sequence,
alt.haplotypes)
# Create edges connecting nodes along a haplotype (or allele in alleleGraph mode)
for haplo in self.get_haplotypes():
nodes = graph.nodes_by_haplo(haplo)
logging.info(
f"Linking nodes in sequence {haplo}:\t{', '.join(n['name'] for n in nodes)}"
)
pnode = None
for node in nodes:
if pnode:
if pnode["end"] == node["start"] - 1:
graph.add_edge(pnode, node, [haplo])
pnode_is_ref_dummy = pnode[
"end"] == pnode["start"] - 1 and not pnode["sequence"]
pnode_ends_before_node = pnode["end"] < node[
"start"] and pnode["start"] < node["start"]
if not pnode_is_ref_dummy and not pnode_ends_before_node:
raise Exception(
f"Inconsistent nodes for haplotype {haplo}: {pnode['name']}, {node['name']}"
)
pnode = node
# In alleleGraph mode link each alt node to all neighboring nodes
# In haplotype mode link nodes without in/out edges to reference
for node in graph.altNodes():
if allele_graph or not any(graph.inEdges(node)):
graph.add_edge(
graph.refNode_ending_at[node["chrom"], node["start"] - 1],
node)
if not any(graph.outEdges(node)):
graph.add_edge(
node, graph.refNode_starting_at[node["chrom"],
node["end"] + 1])
if allele_graph:
isInsertion = node["end"] < node["start"]
for n in graph.nodes_starting_at[node["end"] + 1]:
# Don't loop by connecting multiple insertions at the same position
if not (isInsertion and n["end"] < n["start"]):
graph.add_edge(node, n)
# For nodes that do not have determined in/out edges for a given haplotype
# label all in/out edges as compatible with that haplotype
# excluding edges that connect to another allele at the same vcfVariant (e.g. insertions)
for haplo in self.get_haplotypes():
for node in graph.nodes_by_haplo(haplo):
if not any(graph.inEdges(node, haplo)):
for e in graph.inEdges(node):
graph.add_edge(graph.nodes[e["from"]], node, [haplo])
assert any(graph.inEdges(node, haplo))
if not any(graph.outEdges(node, haplo)):
for e in graph.outEdges(node):
graph.add_edge(node, graph.nodes[e["to"]], [haplo])
return graph