def test_calculate_first_second_third_moment(self, values, moments):
"""
Test static method that calculates the first three moments of a distribution.
"""
feature = SubpocketsFeature()
moments_calculated = feature.calculate_first_second_third_moments(values)
if len(values) > 0 and not all(np.isnan(values)):
assert pytest.approx(moments_calculated, abs=1e-6) == moments
else:
assert all(moments_calculated) == all(moments)
def _get_spatial_features_dict(self, pocket_df):
"""
Get spatial features (distances and moments).
Parameters
----------
pocket_df : kissim.io.PocketDataFrame
DataFrame-based pocket object.
Returns
-------
dict of list of float
Per-subpocket feature values [and coordinates] (values) for distances and moments
[and subpocket centers] (keys).
"""
# Set up spatial features
features = {}
# Add subpockets features
feature = SubpocketsFeature.from_pocket(pocket_df)
features["distances"] = feature._distances
features["moments"] = feature._moments
features["subpocket_centers"] = feature._subpocket_centers
return features
def test_values(self, structure_klifs_id, klifs_session):
"""
Test class property: values.
"""
pocket = PocketDataFrame.from_structure_klifs_id(
structure_klifs_id, klifs_session=klifs_session
)
feature = SubpocketsFeature.from_pocket(pocket)
assert isinstance(feature.values, dict)
assert feature.values == feature._moments
def test_add_subpockets(self, structure_klifs_id, klifs_session, subpockets, subpocket_center):
"""
Test if subpockets are added correctly.
"""
pocket = PocketDataFrame.from_structure_klifs_id(
structure_klifs_id, klifs_session=klifs_session
)
feature = SubpocketsFeature()
pocket = feature._add_subpockets(pocket, subpockets)
assert pocket.subpockets.columns.to_list() == [
"subpocket.name",
"subpocket.color",
"subpocket.center",
]
assert pocket.subpockets["subpocket.name"].to_list() == subpockets["subpocket.name"]
assert pocket.subpockets["subpocket.color"].to_list() == subpockets["subpocket.color"]
subpocket_center_calculated = pocket.subpockets["subpocket.center"][0]
for i, j in zip(subpocket_center_calculated, subpocket_center):
assert pytest.approx(i, abs=1e-4) == j
def test_from_pocket(self, structure_klifs_id, klifs_session):
"""
Test if SubpocketsFeature can be set from a Pocket object.
Test object attribues.
"""
pocket = PocketDataFrame.from_structure_klifs_id(
structure_klifs_id, klifs_session=klifs_session
)
feature = SubpocketsFeature.from_pocket(pocket)
assert isinstance(feature, SubpocketsFeature)
# Test class attributes
# Note: _distances and _moments are tested separately below
assert feature.name == structure_klifs_id
for residue_id, residue_ix in zip(feature._residue_ids, feature._residue_ixs):
if residue_id is not None:
assert isinstance(residue_id, int)
assert isinstance(residue_ix, int)
def test_calculate_distance_to_center(
self, structure_klifs_id, klifs_session, residue_id, subpocket_center, distance
):
"""
Test calculation of distances between a subpocket center and a pocket residues.
"""
pocket = PocketDataFrame.from_structure_klifs_id(
structure_klifs_id, klifs_session=klifs_session
)
feature = SubpocketsFeature.from_pocket(pocket)
ca_atoms = pocket.ca_atoms
distance_calculated = feature._calculate_distance_to_center(
ca_atoms, residue_id, subpocket_center
)
if np.isnan(distance):
assert np.isnan(distance_calculated)
else:
assert pytest.approx(distance_calculated, abs=1e-6) == distance
def test_calculate_distances_to_center(
self, structure_klifs_id, klifs_session, subpocket_center, mean_distance
):
"""
Test calculation of distances between a subpocket center and all pocket residues.
Test also the case that there is no subpocket center.
"""
pocket = PocketDataFrame.from_structure_klifs_id(
structure_klifs_id, klifs_session=klifs_session
)
feature = SubpocketsFeature.from_pocket(pocket)
distances_calculated = feature._calculate_distances_to_center(pocket, subpocket_center)
if subpocket_center is None:
assert len(distances_calculated) == len(pocket._residue_ids)
assert all(np.isnan(distances_calculated))
else:
mean_distance_calculated = np.nanmean(np.array(distances_calculated))
assert pytest.approx(mean_distance_calculated, abs=1e-6) == mean_distance
def test_calculate_distances_and_moments(
self,
structure_klifs_id,
klifs_session,
distances_mean,
moments_mean,
subpocket_centers_mean,
):
"""
Test calculation of distances and moments for all subpockets.
We are testing here the class attributes _distances and _moments, whose values are the
return values from the class methods calculate_distance and calculate_moments.
"""
pocket = PocketDataFrame.from_structure_klifs_id(
structure_klifs_id, klifs_session=klifs_session
)
feature = SubpocketsFeature.from_pocket(pocket)
# Test distances
for _, distances in feature._distances.items():
assert isinstance(distances, list)
distances_mean_calculated = {
name: np.nanmean(distances) for name, distances in feature._distances.items()
}
assert pytest.approx(distances_mean_calculated, abs=1e-6) == distances_mean
# Test moments
for _, moments in feature._moments.items():
assert isinstance(moments, list)
moments_mean_calculated = {
name: np.nanmean(moments) for name, moments in feature._moments.items()
}
assert pytest.approx(moments_mean_calculated, abs=1e-6) == moments_mean
# Test subpocket centers
for _, subpocket_center in feature._subpocket_centers.items():
assert isinstance(subpocket_center, list)
subpocket_centers_mean_calculated = {
name: (np.nanmean(coordinates) if coordinates is not None else None)
for name, coordinates in feature._subpocket_centers.items()
}
assert pytest.approx(subpocket_centers_mean_calculated, abs=1e-6) == subpocket_centers_mean
def test_details(self, structure_klifs_id, klifs_session):
"""
Test class property: details.
"""
pocket = PocketDataFrame.from_structure_klifs_id(
structure_klifs_id, klifs_session=klifs_session
)
feature = SubpocketsFeature.from_pocket(pocket)
assert isinstance(feature.details, dict)
# Distances
assert isinstance(feature.details["distances"], pd.DataFrame)
assert feature.details["distances"].index.to_list() == feature._residue_ixs
# Test first column name and dtypes (the other columns depend on the input subpocket names)
assert feature.details["distances"].columns.to_list()[0] == "residue.id"
assert feature.details["distances"].dtypes[0] == "Int32"
# Moments
assert isinstance(feature.details["moments"], pd.DataFrame)
assert feature.details["moments"].index.to_list() == [1, 2, 3]