def load_json(json) -> GraphContainer:
"""
Construct graph object from JSON representation
:param json: Dictionary of JSON file contents
"""
graph = GraphContainer()
for node in json["nodes"]:
seqs = node.get("sequences", ())
if "reference" in node:
chrom, start, end = parse_region(node["reference"])
graph.add_refNode(chrom, start, end, seqs, node["name"])
elif "position" in node:
chrom, start, end = parse_region(node["position"])
graph.add_altNode(chrom, start, end, node["sequence"], seqs,
node["name"])
else:
graph.nodes[node["name"]] = node
for edge in json["edges"]:
seqs = edge.get("sequences", ())
graph.add_edge(graph.nodes[edge["from"]], graph.nodes[edge["to"]],
seqs)
graph.name = json["model_name"]
graph.paths = json.get("paths", [])
graph.target_regions = json.get("target_regions", [])
graph.check()
return graph
def convert_vcf(vcf,
ref,
target_regions=None,
ref_node_padding=150,
ref_node_max_length=1000,
allele_graph=False,
simplify=True,
alt_paths=False,
alt_splitting=False):
"""
Convert a single VCF file to a graph dictionary
:param vcf: file name of the VCF file
:param ref: reference FASTA file name
:param target_regions: target region list
:param ref_node_padding: padding / read length
:param ref_node_max_length: maximum length before splitting a reference node
:param allele_graph: add edges between any compatible allele pair, not just haplotypes from input
:param simplify: simplify the graph
:param alt_paths: Add all possible non-reference paths to the graph
:param alt_splitting: also split long alt nodes (e.g. long insertions)
:return: dictionary containing JSON graph
"""
graph = GraphContainer("Graph from %s" % vcf)
indexed_vcf = tempfile.NamedTemporaryFile(delete=False, suffix=".vcf.gz")
try:
indexed_vcf.close()
# noinspection PyUnresolvedReferences
pysam.bcftools.view(vcf, "-o", indexed_vcf.name, "-O", "z", catch_stdout=False) # pylint: disable=no-member
# noinspection PyUnresolvedReferences
pysam.bcftools.index(indexed_vcf.name) # pylint: disable=no-member
regions = map(parse_region, target_regions) if target_regions else [(None,)*3]
for (chrom, start, end) in regions:
if chrom is not None:
logging.info(f"Starting work on region: {chrom}:{start}-{end}")
try:
vcfGraph = VCFGraph.create_from_vcf(
ref, indexed_vcf.name, chrom, start, end, ref_node_padding, allele_graph)
except NoVCFRecordsException:
logging.info(f"Region {chrom}:{start}-{end} has no VCF records, skipping.")
continue
logging.info(f"CONSTRUCTED VCF GRAPH:\n{str(vcfGraph)}")
chromGraph = vcfGraph.get_graph(allele_graph)
if ref_node_max_length:
graphUtils.split_ref_nodes(chromGraph, ref_node_max_length, ref_node_padding)
if alt_splitting:
graphUtils.split_alt_nodes(chromGraph, ref_node_max_length, ref_node_padding)
if simplify:
graphUtils.remove_empty_nodes(chromGraph)
graphUtils.combine_nodes(chromGraph)
# Disable edge label simplification for now. May use node-label short-cut later
# graphUtils.remove_redundant_edge_labels(graph)
chromGraph.check()
graphUtils.add_graph(graph, chromGraph)
finally:
os.remove(indexed_vcf.name)
graph.target_regions = target_regions or graph.get_reference_regions()
graphUtils.add_source_sink(graph)
graphUtils.add_ref_path(graph)
if alt_paths:
graphUtils.add_alt_paths(graph)
graph.check()
return graph.json_dict()
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