def bootstrap_worker(params):
"""Worker function to compute boostrap replicates of datasets and indiv. loci"""
bootstrap_trees = []
rep, fullpth, oneliners = params
pth, exe = os.path.split(fullpth)
# first, resample w/ replacement/bootstrap across loci
multilocus_bstrap = get_bootstraps(oneliners)
# Resample w/ replacement/boostrap bases within loci
bootstraps = get_bootstrap_replicates(multilocus_bstrap)
for oneliner in bootstraps:
args_dict, locus = split_oneliner(oneliner, default_model=True)
phylip = oneliner_to_phylip(locus)
phyml = Phyml(phylip, pth=pth, exe=exe)
# run phyml. if no model, defaults to GTR
# TOOD: Why do we need LnL?
args_dict['lnL'], tree = phyml.run(args_dict['model'])
#bootstrap_trees.append("tree '%s' = [&U] %s" % (make_tree_name(args_dict), tree))
bootstrap_trees.append('''%s\t"%s"''' % (rep, tree))
sys.stdout.write("[Info] {0} {1} bootstrap completed\n".format(
strftime("%a, %d %b %Y %H:%M:%S", localtime()),
args_dict['chrm']
)
)
sys.stdout.flush()
return bootstrap_trees
def get_bootstrap_replicates(multilocus_bstrap):
"""Boostrap the bases of all alignments in a resampled population of alignments"""
for locus in multilocus_bstrap:
# split keys from alignments
args_dict, locus = split_oneliner(locus)
# convert alignments to arrays
taxa, numpy_alignment = oneliner_to_array(locus.rstrip(';'))
# transpose so we bootstrap by columns
bases_by_col = np.column_stack(numpy_alignment)
# bootstrap by columns
shuffled = get_bootstraps(bases_by_col)
# transpose back to rows of sequences
shuffled = np.column_stack(shuffled)
# convert back to oneliner
oneliner = array_to_oneliner(taxa, shuffled)
oneliner = "%s:%s" % (make_tree_name(args_dict), oneliner)
yield oneliner
def get_bootstrap_replicates(multilocus_bstrap):
"""Boostrap the bases of all alignments in a resampled population of alignments"""
for locus in multilocus_bstrap:
# split keys from alignments
args_dict, locus = split_oneliner(locus)
# convert alignments to arrays
taxa, numpy_alignment = oneliner_to_array(locus.rstrip(';'))
# transpose so we bootstrap by columns
bases_by_col = np.column_stack(numpy_alignment)
# bootstrap by columns
shuffled = get_bootstraps(bases_by_col)
# transpose back to rows of sequences
shuffled = np.column_stack(shuffled)
# convert back to oneliner
oneliner = array_to_oneliner(taxa, shuffled)
oneliner = "%s:%s" % (make_tree_name(args_dict), oneliner)
yield oneliner
def bootstrap_worker(params):
"""Worker function to compute boostrap replicates of datasets and indiv. loci"""
bootstrap_trees = []
rep, fullpth, oneliners = params
pth, exe = os.path.split(fullpth)
# first, resample w/ replacement/bootstrap across loci
multilocus_bstrap = get_bootstraps(oneliners)
# Resample w/ replacement/boostrap bases within loci
bootstraps = get_bootstrap_replicates(multilocus_bstrap)
for oneliner in bootstraps:
args_dict, locus = split_oneliner(oneliner, default_model=True)
phylip = oneliner_to_phylip(locus)
phyml = Phyml(phylip, pth=pth, exe=exe)
# run phyml. if no model, defaults to GTR
# TOOD: Why do we need LnL?
args_dict['lnL'], tree = phyml.run(args_dict['model'])
#bootstrap_trees.append("tree '%s' = [&U] %s" % (make_tree_name(args_dict), tree))
bootstrap_trees.append('''%s\t"%s"''' % (rep, tree))
sys.stdout.write("[Info] {0} bootstrap completed\n".format(
strftime("%a, %d %b %Y %H:%M:%S", localtime())))
sys.stdout.flush()
return bootstrap_trees
