def test_sklearn_affinity_propagation(self, ens1):
pytest.importorskip('sklearn')
cc1 = encore.cluster([ens1])
cc2 = encore.cluster([ens1], method=encore.AffinityPropagation())
assert len(cc1) == len(cc2), \
"Native and sklearn implementations of affinity "\
"propagation don't agree: mismatch in number of "\
"clusters: {0} {1}".format(len(cc1), len(cc2))
def test_clustering_AffinityPropagation_direct(self, ens1):
pytest.importorskip('sklearn')
method = encore.AffinityPropagation()
distance_matrix = encore.get_distance_matrix(ens1)
cluster_assignment, details = method(distance_matrix)
expected_value = 7
assert len(set(cluster_assignment)) == expected_value, \
"Unexpected result: {0}".format(cluster_assignment)
def test_sklearn_affinity_propagation(self):
cc1 = encore.cluster([self.ens1])
cc2 = encore.cluster([self.ens1], method=encore.AffinityPropagation())
assert_equal(len(cc1),
len(cc2),
err_msg="Native and sklearn implementations of affinity "
"propagation don't agree: mismatch in number of "
"clusters: {0} {1}".format(len(cc1), len(cc2)))
def affinity_propagation(self, preference=-6.0):
'''Performing Affinity Propagation clustering of AMOEBA-run Trp-Cage folding trajectory:-
Default parameter values from MDAnalysis - damping=0.9, max_iter=500, convergence_iter=50
Preference reduced to -10 from -1 to reflect local homogenity within the trajectory'''
print("Performing Affinity Propagation with input preference = %f" %
preference)
clust = encore.cluster(self.univ_cut,
method=encore.AffinityPropagation(
preference=preference, verbose=True))
centroids = [cluster.centroid * self.ncut for cluster in clust]
ids = [cluster.id for cluster in clust]
print(
"Clustering complete! - %d clusters formed with average size = %d frames"
% (len(ids), np.average([cluster.size for cluster in clust])))
coords_centroids = np.zeros((len(centroids), 304, 3), dtype=np.float64)
coords_centroids[:, :, :] = [
self.coords_protein[centroids[i], :, :]
for i in np.arange(0, len(centroids))
]
protein = self.univ2.select_atoms("protein")
univ_centroids = mda.Merge(protein)
univ_centroids.load_new(coords_centroids, format=MemoryReader)
ref = mda.Universe("folded.pdb")
nframes = len(univ_centroids.trajectory)
alignment = align.AlignTraj(univ_centroids,
ref,
select='protein and name CA',
in_memory=True,
verbose=True)
alignment.run()
idvscenter = {
'Cluster ID': [ids[i] for i in range(0, len(ids))],
'Centroid Time (ps)':
[centroids[i] * 10 for i in range(0, len(centroids))],
'Cluster Size': [cluster.size for cluster in clust]
}
idtable = pd.DataFrame(data=idvscenter)
#Visualisation representation set for Trp-Cage - will expand to general proteins later
view = nglview.show_mdanalysis(univ_centroids)
view.add_cartoon(selection="protein")
view.add_licorice('TRP')
view.add_licorice('PRO')
view.center(selection='protein', duration=nframes)
view.player.parameters = dict(frame=True)
return view, idtable, univ_centroids
def test_clustering_AffinityPropagation_direct(self):
method = encore.AffinityPropagation()
distance_matrix = encore.get_distance_matrix(self.ens1)
cluster_assignment, details = method(distance_matrix)
expected_value = 7
assert_equal(
len(set(cluster_assignment)),
expected_value,
err_msg="Unexpected result: {0}".format(cluster_assignment))
def test_clustering_two_different_methods(self, ens1):
pytest.importorskip('sklearn')
cluster_collection = encore.cluster(
[ens1],
method=[
encore.AffinityPropagation(preference=-7.5),
encore.DBSCAN(min_samples=2)
])
assert len(cluster_collection[0]) == len(cluster_collection[1]), \
"Unexpected result: {0}".format(cluster_collection)
def test_clustering_two_different_methods(self):
cluster_collection = encore.cluster(
[self.ens1],
method=[
encore.AffinityPropagation(preference=-7.5),
encore.DBSCAN(min_samples=2)
])
print(cluster_collection)
print(cluster_collection)
assert_equal(
len(cluster_collection[0]),
len(cluster_collection[1]),
err_msg="Unexpected result: {0}".format(cluster_collection))