def test_fpc_hierarch_tm(self):
os.chdir(self.thisd) # Need as otherwise tests that happen in other directories change os.cwd()
mdir = os.path.join(self.testfiles_dir, "models")
models = glob.glob(mdir + os.sep + "*.pdb")
wdir = 'fpc_test'
if not os.path.isdir(wdir):
os.mkdir(wdir)
fpc = fast_protein_cluster.FPC()
num_clusters = 1
score_type = 'tm'
cluster_method = 'hcomplete'
clusters, cluster_data = fpc.cluster(
models=models,
num_clusters=num_clusters,
score_type=score_type,
cluster_method=cluster_method,
work_dir=wdir,
fpc_exe=self.fpc_exe,
nproc=4,
benchmark=True,
)
self.assertEqual(len(clusters), num_clusters)
d = cluster_data[0]
self.assertEqual(d['cluster_num_models'], 16)
self.assertEqual(d['cluster_method'], 'hcomplete_tm')
self.assertEqual(os.path.basename(d['cluster_centroid']), '5_S_00000005.pdb')
shutil.rmtree(wdir)
def cluster_models(
self, models=None, cluster_method=SPICKER_RMSD, num_clusters=1, cluster_dir=None, max_cluster_size=200
):
"""Wrapper function to run clustering of models dependent on the method"""
if len(models) < 2:
raise RuntimeError("Cannot cluster fewer than 2 models!")
logger.info('Generating %d clusters using method: %s', num_clusters, cluster_method)
if cluster_method != 'import' and not len(models):
raise RuntimeError("Cannot find any models for ensembling!")
# Get the cluster_method_type and cluster_score_type from the cluster_method
cluster_method_type, cluster_score_type, cluster_exe = self.parse_cluster_method(cluster_method)
# Set directory
if cluster_method_type not in ['skip']:
pass
elif cluster_method_type in ['import', 'random']:
if not os.path.isdir(cluster_dir):
raise RuntimeError('Cannot find cluster directory: {0}'.format(cluster_dir))
else:
cluster_dir = os.path.join(self.work_dir, 'clustering')
if cluster_method_type == 'fast_protein_cluster':
SCORE_TYPE = 'rmsd'
CLUSTER_METHOD = 'kmeans'
logger.info(
'Running fast_protein_cluster with: score_type: %s and cluster_method: %s', SCORE_TYPE, CLUSTER_METHOD
)
clusters = fast_protein_cluster.FPC().fpc.cluster(
cluster_method=CLUSTER_METHOD,
fpc_exe=cluster_exe,
max_cluster_size=max_cluster_size,
models=models,
num_clusters=num_clusters,
nproc=self.nproc,
score_type=SCORE_TYPE,
work_dir=cluster_dir,
)
elif cluster_method_type == 'import':
clusters = cluster_util.import_cluster(models)
elif cluster_method_type == 'random':
clusters = cluster_util.random_cluster(cluster_method_type, max_cluster_size, models, num_clusters)
elif cluster_method_type == 'spicker':
logger.info('* Running SPICKER to cluster models *')
spickerer = spicker.Spickerer(spicker_exe=cluster_exe)
clusters = spickerer.cluster(
models,
num_clusters=num_clusters,
max_cluster_size=max_cluster_size,
score_type=cluster_score_type,
run_dir=cluster_dir,
score_matrix=None,
nproc=self.nproc,
)
logger.debug(spickerer.results_summary())
else:
raise RuntimeError('Unrecognised clustering method: {}'.format(cluster_method_type))
return clusters