def run_analytical_fit(boot_collection,
ref_collection,
ref_coords,
task=_fast_geo,
rooted=False,
**kwargs):
fit = np.empty((len(boot_collection), ref_coords.shape[1]))
if ISPY3:
query_trees = [
PhyloTree(tree.encode(), rooted) for tree in boot_collection.trees
]
ref_trees = [
PhyloTree(tree.encode(), rooted) for tree in ref_collection.trees
]
else:
query_trees = [
PhyloTree(tree, rooted) for tree in boot_collection.trees
]
ref_trees = [PhyloTree(tree, rooted) for tree in ref_collection.trees]
for i, tree in enumerate(query_trees):
ref_dists = np.array(
[task(tree, ref_tree, False) for ref_tree in ref_trees])
aft = AnalyticalFit(ref_coords.values, **kwargs)
fit[i] = aft.fit(ref_dists)
return fit
def run_out_of_sample_mds(boot_collection,
ref_collection,
ref_distance_matrix,
index,
dimensions,
task=_fast_geo,
rooted=False,
**kwargs):
"""
index = index of the locus the bootstrap sample corresponds to - only important if
using recalc=True in kwargs
"""
fit = np.empty((len(boot_collection), dimensions))
if ISPY3:
query_trees = [
PhyloTree(tree.encode(), rooted) for tree in boot_collection.trees
]
ref_trees = [
PhyloTree(tree.encode(), rooted) for tree in ref_collection.trees
]
else:
query_trees = [
PhyloTree(tree, rooted) for tree in boot_collection.trees
]
ref_trees = [PhyloTree(tree, rooted) for tree in ref_collection.trees]
for i, tree in enumerate(query_trees):
distvec = np.array(
[task(tree, ref_tree, False) for ref_tree in ref_trees])
oos = OutOfSampleMDS(ref_distance_matrix)
fit[i] = oos.fit(index, distvec, dimensions=dimensions, **kwargs)
return fit
def run_optimise_bootstrap_coords(boot_collection,
ref_collection,
ref_coords,
task=_fast_geo,
rooted=False,
**kwargs):
fit = np.empty((len(boot_collection), ref_coords.shape[1]))
if ISPY3:
query_trees = [
PhyloTree(tree.encode(), rooted) for tree in boot_collection.trees
]
ref_trees = [
PhyloTree(tree.encode(), rooted) for tree in ref_collection.trees
]
else:
query_trees = [
PhyloTree(tree, rooted) for tree in boot_collection.trees
]
ref_trees = [PhyloTree(tree, rooted) for tree in ref_collection.trees]
for i, tree in enumerate(query_trees):
ref_dists = np.array(
[task(tree, ref_tree, False) for ref_tree in ref_trees])
opt = OptimiseDistanceFit(ref_coords.values, ref_dists)
fit[i] = opt.newton(**kwargs)
return fit
def get_inter_tree_distances(self,
metric,
jobhandler=default_jobhandler,
normalise=False,
batchsize=1):
""" Generate a distance matrix from a fully-populated Collection """
metrics = {
'euc': tasks.EuclideanTreeDistance,
'geo': tasks.GeodesicTreeDistance,
'rf': tasks.RobinsonFouldsTreeDistance,
'wrf': tasks.WeightedRobinsonFouldsTreeDistance,
'fasteuc': tasks.EqualLeafSetEuclideanTreeDistance,
'fastgeo': tasks.EqualLeafSetGeodesicTreeDistance,
'fastrf': tasks.EqualLeafSetRobinsonFouldsTreeDistance,
'fastwrf': tasks.EqualLeafSetWeightedRobinsonFouldsTreeDistance
}
optioncheck(metric, list(metrics.keys()))
task_interface = metrics[metric]()
if metric.startswith('fast'):
trees = (PhyloTree(newick, False) for newick in self.trees)
else:
trees = self.trees
args = task_interface.scrape_args(trees, normalise)
logger.debug('{}'.format(args))
msg = task_interface.name
array = jobhandler(task_interface.get_task(), args, msg, batchsize)
return DistanceMatrix.from_array(squareform(array), self.names)
def phylotree(self):
"""
Gets the c++ PhyloTree object corresponding to this tree.
Should be canonically the same - we set a _dirty flag if the Python version of the
tree has changed since construction. If the flag is set then we reconstruct
the c++ PhyloTree
:return: PhyloTree instance
"""
if not self._phylotree or self._dirty:
self._phylotree = PhyloTree(self.newick, self.rooted)
self._dirty = False
return self._phylotree
def get_inter_tree_distances(self, metric, jobhandler=default_jobhandler,
normalise=False, min_overlap=4, overlap_fail_value=0,
batchsize=1, show_progress=True):
""" Generate a distance matrix from a fully-populated Collection.
Can silence progressbars with show_progress=False option
:param metric: str. Tree distance metric to use. Choice of 'euc', 'geo', 'rf', 'wrf'.
:param jobhandler: treeCl.Jobhandler. Choice of SequentialJobHandler, ThreadpoolJobHandler, or
ProcesspoolJobHandler.
:param normalise: Bool. Whether to normalise the tree distance to the size of the leaf set.
:param min_overlap: int. Trees with fewer leaves in common than this threshold will not have their distance
calculated, but instead the distance returned will be the value in `overlap_fail_value`.
:param overlap_fail_value: Any. The distance between trees with fewer leaves in common than `min_overlap`
is set to this value.
:param batchsize: int. Number of jobs to process in a batch when using a ProcesspoolJobHandler or a
ThreadpoolJobHandler.
:return: treeCl.DistanceMatrix.
"""
metrics = {'euc': tasks.EuclideanTreeDistance,
'geo': tasks.GeodesicTreeDistance,
'rf': tasks.RobinsonFouldsTreeDistance,
'wrf': tasks.WeightedRobinsonFouldsTreeDistance,
'fasteuc': tasks.EqualLeafSetEuclideanTreeDistance,
'fastgeo': tasks.EqualLeafSetGeodesicTreeDistance,
'fastrf': tasks.EqualLeafSetRobinsonFouldsTreeDistance,
'fastwrf': tasks.EqualLeafSetWeightedRobinsonFouldsTreeDistance}
optioncheck(metric, list(metrics.keys()))
task_interface = metrics[metric]()
if metric.startswith('fast'):
trees = (PhyloTree(newick, False) for newick in self.trees)
else:
trees = self.trees
args = task_interface.scrape_args(trees, normalise, min_overlap, overlap_fail_value)
logger.debug('{}'.format(args))
msg = task_interface.name if show_progress else ''
array = jobhandler(task_interface.get_task(), args, msg, batchsize, nargs=binom_coeff(len(trees)))
return DistanceMatrix.from_array(squareform(array), self.names)