def get_backbone_tree(tree1, tree2):
"""Constain tree1 to its shared leaf set with tree2
Parameters
----------
tree1 : dendropy tree object
tree2 : dendropy tree object
Returns
-------
tree1 : dendropy tree object
"""
tree1 = deepcopy(tree1)
leaves1 = njmergepair.get_leaf_set(tree1)
leaves2 = njmergepair.get_leaf_set(tree2)
shared = list(leaves1.intersection(leaves2))
taxa = dendropy.TaxonNamespace(shared)
tree1.retain_taxa_with_labels(shared)
tree1.migrate_taxon_namespace(taxa)
tree1.encode_bipartitions()
return tree1
def map_splits_to_node_list(big_tree, lil_tree):
"""Map a split in little tree to a list of nodes in big tree
NOTE: Because little tree is *contained* within big tree, more
than one node in big tree can be mapped to the same split.
Parameters
----------
big_tree : dendropy tree object
lil_tree : dendropy tree object
Returns
-------
split_to_node_list : python dictionary
keys - splits encoded as integers (read below!)
values - lists of nodes in a dendropy tree object
"""
lil_leafset = njmergepair.get_leaf_set(lil_tree)
split_to_node_list = {}
for node in big_tree.postorder_node_iter():
big_leafset = njmergepair.get_leaf_set(node)
shared_leafset = list(big_leafset.intersection(lil_leafset))
if len(shared_leafset) > 0:
split = lil_tree.taxon_namespace.taxa_bitmask(
labels=shared_leafset)
try:
split_to_node_list[split].append(node)
except KeyError:
split_to_node_list[split] = []
split_to_node_list[split].append(node)
return split_to_node_list
def dscmcombine(workdir, trees, mstmat, outfile):
"""
Parameters
----------
workdir : str
working or output directory
Results
-------
Nothing
"""
# Turn matrix into graph
graph = networkx.Graph(mstmat)
# Merge trees
combined_tree = None
nodes = list(deepcopy(graph.nodes()))
root = None
next_roots = [nodes[0]]
while len(nodes) > 0:
if root is None:
root = next_roots[0]
neighbors = graph.neighbors(root)
next_roots = list(set(next_roots).union(set(neighbors)))
if len(neighbors) == 0:
nodes.remove(root)
next_roots.remove(root)
root = None
else:
sys.stdout.write("Combining %d and %d...\n" % (root, neighbors[0]))
if root < neighbors[0]:
i = root
j = neighbors[0]
else:
i = neighbors[0]
j = root
tijfile = name_treepair_file(workdir, trees[i], trees[j])
if not os.path.exists(tijfile):
tijfile = name_treepair_file(workdir, trees[j], trees[i])
if combined_tree is None:
combined_tree = dendropy.Tree.get(path=tijfile,
schema="newick")
else:
tij = dendropy.Tree.get(path=tijfile, schema="newick")
combine_two_trees_via_dscm(tij, combined_tree)
combined_tree.update_bipartitions()
sys.stdout.write("...combined tree has %d leaves!\n" %
len(njmergepair.get_leaf_set(combined_tree)))
graph.remove_edge(root, neighbors[0])
with open(outfile, 'w') as f:
f.write(combined_tree.as_string(schema="newick")[5:])
def combine_two_trees_via_dscm(tree_AB, tree_BC):
"""Combines two trees via distance-based strict consensus merger
A is the subset of leaves only in tree AB
C is the subset of leaves only in tree BC
B is the subset of leaves in both tree AB and tree BC
Tree AB and tree BC must be equivalent on their shared leaf set B
Parameters
----------
tree_AB : dendropy tree object
tree_BC : dendropy tree object
Returns
-------
tree_BC : dendropy tree object
tree BC with leaves from set A added
"""
data = list(
njmergepair.get_leaf_set(tree_AB).intersection(
njmergepair.get_leaf_set(tree_BC)))
taxa = dendropy.TaxonNamespace(data)
# [incompatible] = are_two_trees_incompatible(tree_AB, tree_BC)
# if incompatible:
# sys.exit("Input trees are not compatible!\n")
backbone_tree = get_backbone_tree(tree_AB, tree_BC)
if backbone_tree is None:
raise Exception("Unable to extract a backbone tree!\n")
# Root all trees at the same shared leaf --
# required for split mapping and post-order traversal to work!
root = backbone_tree.taxon_namespace[0].label
node_XX = backbone_tree.find_node_with_taxon_label(root)
node_AB = tree_AB.find_node_with_taxon_label(root)
node_BC = tree_BC.find_node_with_taxon_label(root)
elen_AB = node_AB.edge.length / 2.0
elen_BC = node_BC.edge.length / 2.0
backbone_tree.is_rooted = True
tree_AB.is_rooted = True
tree_BC.is_rooted = True
backbone_tree.reroot_at_edge(node_XX.edge)
tree_AB.reroot_at_edge(node_AB.edge)
tree_BC.reroot_at_edge(node_BC.edge)
node_AB.edge.length = elen_AB
node_BC.edge.length = elen_BC
node_AB.sibling_nodes()[0].edge.length = elen_AB
node_BC.sibling_nodes()[0].edge.length = elen_BC
# Map nodes based on splits in shared leaf set
map_AB = map_splits_to_node_list(tree_AB, backbone_tree)
map_BC = map_splits_to_node_list(tree_BC, backbone_tree)
# Add missing taxa from AB **into** BC using the
# distance-based SCM strategy to handle collisions
nodes = [n for n in backbone_tree.postorder_node_iter()]
for node in nodes[:-1]:
clade = njmergepair.get_leaf_set(node)
split = backbone_tree.taxon_namespace.taxa_bitmask(labels=clade)
node_path_AB = map_AB[split]
node_path_BC = map_BC[split]
num_edges_AB = len(node_path_AB)
num_edges_BC = len(node_path_BC)
sibs_path_AB = []
for n in node_path_AB:
s = n.sibling_nodes()
if len(s) > 1:
sys.exit("Tree AB is not binary!\n")
sibs_path_AB.append(s[0])
sibs_path_BC = []
for n in node_path_BC:
s = n.sibling_nodes()
if len(s) > 1:
sys.exit("Tree BC is not binary!\n")
sibs_path_BC.append(s[0])
if num_edges_AB > 1 and num_edges_BC > 1:
# Found a collision -- add edges from tree AB to tree BC
# Find normalization factor for path in AB
# ALSO compute the point of attachment for edges in the path
# IN ORDER to identify the ORDER in which edges should be added
elen_AB = node_path_AB[0].edge.length
elen_path_AB = [elen_AB]
i_AB = 1
for node_AB in node_path_AB[1:]:
elen_AB = elen_AB + node_AB.edge.length
elen_path_AB.append(elen_path_AB[i_AB - 1] +
node_AB.edge.length)
i_AB = i_AB + 1
elen_BC = node_path_BC[0].edge.length
elen_path_BC = [elen_BC]
i_BC = 1
for node_BC in node_path_BC[1:]:
elen_BC = elen_BC + node_BC.edge.length
elen_path_BC.append(elen_path_BC[i_BC - 1] +
node_BC.edge.length)
i_BC = i_BC + 1
if elen_AB == 0.0 or elen_BC == 0.0:
raise Exception("Collision on path of length zero!\n")
norm_AB = elen_BC / elen_AB
for i in range(len(elen_path_AB)):
elen_path_AB[i] = elen_path_AB[i] * norm_AB
# Extract components of tree BC
node_BC = node_path_BC[-1]
sibs_BC = sibs_path_BC[-1]
parent_BC = node_BC.parent_node
parent_BC.clear_child_nodes()
# Get node in tree BC and update branch length
child1 = node_path_BC[0]
elen_AB = node_path_AB[0].edge.length * norm_AB
elen_BC = node_path_BC[0].edge.length
if elen_AB < elen_BC:
child1.edge.length = elen_AB
else:
child1.edge.length = elen_BC
i_AB = 0
i_BC = 0
start = None
while (True):
dothis = None
if i_AB < num_edges_AB - 1 and i_BC < num_edges_BC - 1:
if elen_path_AB[i_AB] == elen_path_BC[i_BC]:
# Add a new node created from AB and BC
dothis = 3
elif elen_path_AB[i_AB] < elen_path_BC[i_BC]:
# Add remaining edges from tree AB
dothis = 1
else:
# Add remaining edges from tree BC
dothis = 2
elif i_AB < num_edges_AB - 1:
# Add remaining edges from tree AB
dothis = 1
elif i_BC < num_edges_BC - 1:
# Add remaining edges from tree BC
dothis = 2
else:
# No more edges to add!
break
if dothis == 1:
# Adding AB only
child2 = sibs_path_AB[i_AB]
if start is None:
start = child1.edge.length
else:
stop = elen_path_AB[i_AB]
child1.edge.length = stop - start
start = stop
i_AB = i_AB + 1
elif dothis == 2:
# Add BC only
child2 = sibs_path_BC[i_BC]
if start is None:
start = child1.edge.length
else:
stop = elen_path_BC[i_BC]
child1.edge.length = stop - start
start = stop
i_BC = i_BC + 1
else:
# Add both AB and BC
child2 = dendropy.Node()
child2.set_child_nodes(
[sibs_path_AB[i_AB], sibs_path_BC[i_BC]])
child2.edge.length = 0.0
if start is None:
start = child1.edge.length
else:
stop = elen_path_BC[i_BC]
child1.edge.length = stop - start
start = stop
i_AB = i_AB + 1
i_BC = i_BC + 1
child2.parent_node = None
# Combine nodes from tree BC (node 1) and
# tree AB or tree BC (node 2)
next_node = dendropy.Node()
next_node.set_child_nodes([child1, child2])
# Set node in tree BC as next_node
child1 = next_node
# Recombine the three components of tree BC
next_node.edge.length = elen_path_BC[-1] - start
parent_BC.set_child_nodes([next_node] + [sibs_BC])
elif num_edges_AB > 1:
# Found edges in tree AB not in tree BC --
# add edges from tree AB to tree BC!
# Find normalization factor for path in AB
elen_BC = node_path_BC[0].edge.length
elen_AB = 0.0
elen_path_AB = []
for node_AB in node_path_AB:
elen_AB = elen_AB + node_AB.edge.length
elen_path_AB.append(node_AB.edge.length)
if elen_AB == 0.0:
xxxx_AB = elen_BC / num_edges_AB
for i in range(num_edges_AB):
elen_path_AB[i] = xxxx_AB
else:
norm_AB = elen_BC / elen_AB
for i in range(num_edges_AB):
elen_path_AB[i] = elen_path_AB[i] * norm_AB
# Extract components of tree BC
node_BC = node_path_BC[0]
sibs_BC = sibs_path_BC[0]
parent_BC = node_BC.parent_node
parent_BC.clear_child_nodes()
# Get node in tree BC and update branch length to match tree AB
child1 = node_path_BC[0]
child1.edge.length = elen_path_AB[0]
# Add each nodes in tree AB to tree BC
for i_AB in range(1, num_edges_AB):
# Remove parent from child 1
child1.parent_node = None
# Get node being added from tree AB!
child2 = sibs_path_AB[i_AB - 1]
child2.parent_node = None
# Combine nodes from tree BC (node 1) and tree AB (node 2)
next_node = dendropy.Node()
next_node.set_child_nodes([child1, child2])
next_node.edge.length = elen_path_AB[i_AB]
# Set node in tree BC as next_node
child1 = next_node
# Recombine the three components of tree BC
parent_BC.set_child_nodes([next_node] + [sibs_BC])
elif num_edges_BC > 1:
# Found edges in tree BC not in tree AB
pass
else:
# Found one edge in tree BC and one edge in tree AB
pass
tree_BC.migrate_taxon_namespace(taxa)
