def test_format_weights_valueerror(feature_weights):
"""
Test if wrong data type of input feature weights raises TypeError.
"""
with pytest.raises(ValueError):
format_weights(feature_weights)
def from_feature_distances(cls, feature_distances, feature_weights=None):
"""
Get fingerprint distance.
Parameters
----------
feature_distances : kissim.similarity.FeatureDistances
Distances and bit coverages between two fingerprints for each of their features.
feature_weights : None or list of float
Feature weights of the following form:
(i) None
Default feature weights: All features equally distributed to 1/15
(15 features in total).
(ii) By feature (list of 15 floats):
Features to be set in the following order: size, hbd, hba, charge, aromatic,
aliphatic, sco, exposure, distance_to_centroid, distance_to_hinge_region,
distance_to_dfg_region, distance_to_front_pocket, moment1, moment2, and moment3.
All floats must sum up to 1.0.
Returns
-------
kissim.comparison.FingerprintDistance
Fingerprint distance.
"""
fingerprint_distance = cls()
# Get data of interest from input
weights = format_weights(feature_weights)
bit_coverages = feature_distances.bit_coverages
distances = feature_distances.distances
# Set class attributes
fingerprint_distance.structure_pair_ids = feature_distances.structure_pair_ids
fingerprint_distance.kinase_pair_ids = feature_distances.kinase_pair_ids
# Calculate weighted sum of feature bit coverages
fingerprint_distance.distance = fingerprint_distance._distance(
distances, weights)
# Calculate weighted sum of feature distances
fingerprint_distance.bit_coverage = fingerprint_distance._bit_coverage(
bit_coverages, weights)
return fingerprint_distance
def test_format_weights(feature_weights, feature_weights_formatted):
"""
Test if feature weights are added correctly to feature distance DataFrame.
Parameters
----------
feature_weights : None or list of float
Feature weights.
feature_weights_formatted : list of float
Formatted feature weights of length 15.
"""
feature_weights_formatted_calculated = format_weights(feature_weights)
assert np.isclose(
np.std(feature_weights_formatted),
np.std(feature_weights_formatted_calculated),
rtol=1e-04,
)
def from_feature_distances_generator(cls,
feature_distances_generator,
feature_weights=None):
"""
Generate fingerprint distances for multiple fingerprint pairs based on their feature
distances, given a feature weighting scheme.
Uses parallel computing of fingerprint pairs.
Parameters
----------
feature_distances_generator : kissim.similarity.FeatureDistancesGenerator
Feature distances for multiple fingerprint pairs.
feature_weights : None or list of float
Feature weights of the following form:
(i) None
Default feature weights: All features equally distributed to 1/15
(15 features in total).
(ii) By feature (list of 15 floats):
Features to be set in the following order: size, hbd, hba, charge, aromatic,
aliphatic, sco, exposure, distance_to_centroid, distance_to_hinge_region,
distance_to_dfg_region, distance_to_front_pocket, moment1, moment2, and moment3.
All floats must sum up to 1.0.
Returns
-------
kissim.comparison.FingerprintDistanceGenerator
Fingerprint distance generator.
"""
logger.info("GENERATE FINGERPRINT DISTANCES")
# logger.info(f"Number of input feature distances: {len(feature_distances_generator.data)}")
start_time = datetime.datetime.now()
logger.info(
f"Fingerprint distance generation started at: {start_time}")
# Format input feature weights
feature_weights = format_weights(feature_weights)
logger.info(f"Feature weights: {feature_weights}")
# Weighted sum of pairwise feature distances and bit coverages
fingerprint_distance = FingerprintDistance()
distances = [
fingerprint_distance._distance(distances, feature_weights)
for distances in tqdm(
feature_distances_generator.distances,
desc="Calculate pairwise fingerprint distance",
)
]
bit_coverages = [
fingerprint_distance._bit_coverage(bit_coverages, feature_weights)
for bit_coverages in tqdm(
feature_distances_generator.bit_coverages,
desc="Calculate pairwise fingerprint coverage",
)
]
# Set class attributes
fingerprint_distance_generator = cls()
fingerprint_distance_generator.data = feature_distances_generator.data[
["structure.1", "structure.2", "kinase.1", "kinase.2"]].copy()
fingerprint_distance_generator.data["distance"] = distances
fingerprint_distance_generator.data["bit_coverage"] = bit_coverages
fingerprint_distance_generator.structure_kinase_ids = (
feature_distances_generator.structure_kinase_ids)
logger.info(
f"Number of output fingerprint distances: {len(fingerprint_distance_generator.data)}"
)
end_time = datetime.datetime.now()
logger.info(f"Runtime: {end_time - start_time}")
return fingerprint_distance_generator