def train(self, home_dir=None, sf_pickle=None, pdbbind_version=2016):
if not home_dir:
home_dir = dirname(__file__) + '/RFScore'
desc_path = path_join(home_dir, 'rfscore_descs_v%i.csv' % self.version)
super(rfscore,
self)._load_pdbbind_desc(desc_path,
pdbbind_version=pdbbind_version)
# remove sparse dimentions
if self.spr > 0:
self.mask = (self.train_descs > self.spr).any(axis=0)
if self.mask.sum() > 0:
self.train_descs = self.train_descs[:, self.mask]
self.test_descs = self.test_descs[:, self.mask]
# make nets reproducible
random_seed(1)
self.model.fit(self.train_descs, self.train_target)
print('Training RFScore v%i on PDBBind v%i' %
(self.version, pdbbind_version),
file=sys.stderr)
sets = [
('Test', self.model.predict(self.test_descs), self.test_target),
('Train', self.model.predict(self.train_descs), self.train_target),
('OOB', self.model.oob_prediction_, self.train_target)
]
for name, pred, target in sets:
print('%s set:' % name,
'R2_score: %.4f' % r2_score(target, pred),
'Rp: %.4f' % pearsonr(target, pred)[0],
'RMSE: %.4f' % rmse(target, pred),
'SD: %.4f' % standard_deviation_error(target, pred),
sep='\t',
file=sys.stderr)
# compile trees
if compiledtrees is not None:
try:
print('Compiling Random Forest using sklearn-compiledtrees',
file=sys.stderr)
self.model = compiledtrees.CompiledRegressionPredictor(
self.model, n_jobs=self.n_jobs)
except Exception as e:
print('Failed to compile Random Forest with exception: %s' % e,
file=sys.stderr)
print('Continuing without compiled RF.', file=sys.stderr)
if sf_pickle is None:
return self.save('RFScore_v%i_pdbbind%i.pickle' %
(self.version, pdbbind_version))
else:
return self.save(sf_pickle)
def train(self, home_dir=None, sf_pickle=None, pdbbind_version=2016):
if not home_dir:
home_dir = dirname(__file__) + '/RFScore'
desc_path = path_join(home_dir, 'rfscore_descs_v%i.csv' % self.version)
super(rfscore, self)._load_pdbbind_desc(desc_path,
pdbbind_version=pdbbind_version)
# remove sparse dimentions
if self.spr > 0:
self.mask = (self.train_descs > self.spr).any(axis=0)
if self.mask.sum() > 0:
self.train_descs = self.train_descs[:, self.mask]
self.test_descs = self.test_descs[:, self.mask]
# make nets reproducible
random_seed(1)
self.model.fit(self.train_descs, self.train_target)
print('Training RFScore v%i on PDBBind v%i'
% (self.version, pdbbind_version), file=sys.stderr)
sets = [
('Test', self.model.predict(self.test_descs), self.test_target),
('Train', self.model.predict(self.train_descs), self.train_target),
('OOB', self.model.oob_prediction_, self.train_target)]
for name, pred, target in sets:
print('%s set:' % name,
'R2_score: %.4f' % r2_score(target, pred),
'Rp: %.4f' % pearsonr(target, pred)[0],
'RMSE: %.4f' % rmse(target, pred),
'SD: %.4f' % standard_deviation_error(target, pred),
sep='\t', file=sys.stderr)
# compile trees
if compiledtrees is not None:
try:
print('Compiling Random Forest using sklearn-compiledtrees',
file=sys.stderr)
self.model = compiledtrees.CompiledRegressionPredictor(
self.model, n_jobs=self.n_jobs)
except Exception as e:
print('Failed to compile Random Forest with exception: %s' % e,
file=sys.stderr)
print('Continuing without compiled RF.', file=sys.stderr)
if sf_pickle is None:
return self.save('RFScore_v%i_pdbbind%i.pickle'
% (self.version, pdbbind_version))
else:
return self.save(sf_pickle)
def train(self, home_dir=None, sf_pickle=None, pdbbind_version=2016):
if not home_dir:
home_dir = dirname(__file__) + '/NNScore'
desc_path = path_join(home_dir, 'nnscore_descs.csv')
super(nnscore, self)._load_pdbbind_desc(desc_path,
pdbbind_version=pdbbind_version)
# number of network to sample; original implementation did 1000, but
# 100 give results good enough.
# TODO: allow user to specify number of nets?
n = 1000
# make nets reproducible
random_seed(1)
seeds = np.random.randint(123456789, size=n)
trained_nets = (
Parallel(n_jobs=self.n_jobs, verbose=10, pre_dispatch='all')(
delayed(method_caller)(
neuralnetwork((5,),
random_state=seeds[i],
activation='logistic',
solver='lbfgs',
max_iter=10000),
'fit',
self.train_descs,
self.train_target)
for i in range(n)))
# get 20 best
trained_nets.sort(key=lambda n: n.score(self.test_descs,
self.test_target.flatten()))
self.model = ensemble_model(trained_nets[-20:])
sets = [
('Test', self.model.predict(self.test_descs), self.test_target),
('Train', self.model.predict(self.train_descs), self.train_target)]
for name, pred, target in sets:
if len(target) < 3:
print('There are less than 3 values to predict, skipping.', file=sys.stderr)
continue
print('%s set:' % name,
'R2_score: %.4f' % r2_score(target, pred),
'Rp: %.4f' % pearsonr(target, pred)[0],
'RMSE: %.4f' % rmse(target, pred),
'SD: %.4f' % standard_deviation_error(target, pred),
sep='\t', file=sys.stderr)
if sf_pickle is None:
return self.save('NNScore_pdbbind%i.pickle' % (pdbbind_version))
else:
return self.save(sf_pickle)
def train(self, home_dir=None, sf_pickle=None, pdbbind_version=2016):
if not home_dir:
home_dir = dirname(__file__) + '/NNScore'
desc_path = path_join(home_dir, 'nnscore_descs.csv')
super(nnscore, self)._load_pdbbind_desc(desc_path,
pdbbind_version=pdbbind_version)
# number of network to sample; original implementation did 1000, but
# 100 give results good enough.
# TODO: allow user to specify number of nets?
n = 1000
# make nets reproducible
random_seed(1)
seeds = np.random.randint(123456789, size=n)
trained_nets = (
Parallel(n_jobs=self.n_jobs, verbose=10, pre_dispatch='all')(
delayed(method_caller)(
neuralnetwork((5,),
random_state=seeds[i],
activation='logistic',
solver='lbfgs',
max_iter=10000),
'fit',
self.train_descs,
self.train_target)
for i in range(n)))
# get 20 best
trained_nets.sort(key=lambda n: n.score(self.test_descs,
self.test_target.flatten()))
self.model = ensemble_model(trained_nets[-20:])
sets = [
('Test', self.model.predict(self.test_descs), self.test_target),
('Train', self.model.predict(self.train_descs), self.train_target)]
for name, pred, target in sets:
print('%s set:' % name,
'R2_score: %.4f' % r2_score(target, pred),
'Rp: %.4f' % pearsonr(target, pred)[0],
'RMSE: %.4f' % rmse(target, pred),
'SD: %.4f' % standard_deviation_error(target, pred),
sep='\t', file=sys.stderr)
if sf_pickle is None:
return self.save('NNScore_pdbbind%i.pickle' % (pdbbind_version))
else:
return self.save(sf_pickle)
def train(self, home_dir=None, sf_pickle=None, pdbbind_version=2016,
ignore_json=False):
if not home_dir:
home_dir = path_join(dirname(__file__), 'PLECscore')
desc_path = path_join(home_dir, 'plecscore_descs_p%i_l%i.csv.gz' %
(self.depth_protein, self.depth_ligand))
json_path = path_join(
home_dir, 'plecscore_%s_p%i_l%i_s%i_pdbbind%i.json' %
(self.version, self.depth_protein,
self.depth_ligand, self.size, pdbbind_version))
if (self.version in ['linear'] and # TODO: support other models
isfile(json_path) and
not ignore_json):
print('Loading pretrained PLECscore %s with depths P%i L%i on '
'PDBBind v%i'
% (self.version, self.depth_protein, self.depth_ligand,
pdbbind_version), file=sys.stderr)
with open(json_path) as json_f:
json_data = json.load(json_f)
for k, v in json_data.items():
if isinstance(v, list):
if isinstance(v[0], list):
v = [np.array(x) for x in v]
else:
v = np.array(v)
setattr(self.model, k, v)
else:
# blacklist core set 2013 and astex
pdbids_blacklist = [
'3ao4', '3i3b', '1uto', '1ps3', '1qi0', '3g2z', '3dxg', '3l7b',
'3mfv', '3b3s', '3kgp', '3fk1', '3fcq', '3lka', '3udh', '4gqq',
'3imc', '2xdl', '2ymd', '1lbk', '1bcu', '3zsx', '1f8d', '3muz',
'2v00', '1loq', '3n7a', '2r23', '3nq3', '2hb1', '2w66', '1n2v',
'3kwa', '3g2n', '4de2', '3ozt', '3b3w', '3cft', '3f3a', '2qmj',
'3f80', '1a30', '1w3k', '3ivg', '2jdy', '3u9q', '3pxf', '2wbg',
'1u33', '2x0y', '3mss', '1vso', '1q8t', '3acw', '3bpc', '3vd4',
'3cj2', '2brb', '1p1q', '2vo5', '3d4z', '2gss', '2yge', '3gy4',
'3zso', '3ov1', '1w4o', '1zea', '2zxd', '3ueu', '2qft', '1gpk',
'1f8b', '2jdm', '3su5', '2wca', '3n86', '2x97', '1n1m', '1o5b',
'2y5h', '3ehy', '4des', '3ebp', '1q8u', '4de1', '3huc', '3l4w',
'2vl4', '3coy', '3f3c', '1os0', '3owj', '3bkk', '1yc1', '1hnn',
'3vh9', '3bfu', '1w3l', '3k5v', '2qbr', '1lol', '10gs', '2j78',
'1r5y', '2weg', '3uo4', '3jvs', '2yfe', '1sln', '2iwx', '2jdu',
'4djv', '2xhm', '2xnb', '3s8o', '2zcr', '3oe5', '3gbb', '2d3u',
'3uex', '4dew', '1xd0', '1z95', '2vot', '1oyt', '2ole', '3gcs',
'1kel', '2vvn', '3kv2', '3pww', '3su2', '1f8c', '2xys', '3l4u',
'2xb8', '2d1o', '2zjw', '3f3e', '2g70', '2zwz', '1u1b', '4g8m',
'1o3f', '2x8z', '3cyx', '2cet', '3ag9', '2pq9', '3l3n', '1nvq',
'2cbj', '2v7a', '1h23', '2qbp', '3b68', '2xbv', '2fvd', '2vw5',
'3ejr', '3f17', '3nox', '1hfs', '1jyq', '2pcp', '3ge7', '2wtv',
'2zcq', '2obf', '3e93', '2p4y', '3dd0', '3nw9', '3uri', '3gnw',
'3su3', '2xy9', '1sqa', '3fv1', '2yki', '3g0w', '3pe2', '1e66',
'1igj', '4tmn', '2zx6', '3myg', '4gid', '3utu', '1lor', '1mq6',
'2x00', '2j62', '4djr', '1gm8', '1gpk', '1hnn', '1hp0', '1hq2',
'1hvy', '1hwi', '1hww', '1ia1', '1j3j', '1jd0', '1jje', '1ke5',
'1kzk', '1l2s', '1l7f', '1lpz', '1m2z', '1mmv', '1mzc', '1n1m',
'1n2v', '1n46', '1nav', '1of1', '1of6', '1opk', '1oq5', '1owe',
'1oyt', '1p2y', '1p62', '1pmn', '1q1g', '1q41', '1q4g', '1r1h',
'1r55', '1r58', '1r9o', '1s19', '1s3v', '1sg0', '1sj0', '1sq5',
'1sqn', '1t40', '1t46', '1t9b', '1tow', '1tt1', '1u1c', '1uml',
'1unl', '1uou', '1v0p', '1v48', '1v4s', '1vcj', '1w1p', '1w2g',
'1xm6', '1xoq', '1xoz', '1y6b', '1ygc', '1yqy', '1yv3', '1yvf',
'1ywr', '1z95', '2bm2', '2br1', '2bsm']
# use remote csv if it's not present
if not isfile(desc_path):
branch = 'master' # define branch/commit
desc_url = ('https://raw.githubusercontent.com/oddt/oddt/%s'
'/oddt/scoring/functions/PLECscore/'
'plecscore_descs_p%i_l%i.csv.gz' %
(branch, self.depth_protein, self.depth_ligand))
warnings.warn('The CSV for PLEC P%i L%i is missing. Trying to '
'get it from ODDT GitHub.' % (self.depth_protein,
self.depth_ligand))
# download and save CSV
pd.read_csv(desc_url, index_col='pdbid').to_csv(
desc_path, compression='gzip')
# set PLEC size to unfolded
super(PLECscore, self)._load_pdbbind_desc(
desc_path,
train_set=('general', 'refined'),
pdbbind_version=pdbbind_version,
train_blacklist=pdbids_blacklist,
fold_size=self.size,
)
print('Training PLECscore %s with depths P%i L%i on PDBBind v%i'
% (self.version, self.depth_protein, self.depth_ligand,
pdbbind_version), file=sys.stderr)
self.model.fit(self.train_descs, self.train_target)
sets = [
('Test', self.model.predict(self.test_descs), self.test_target),
('Train', self.model.predict(self.train_descs), self.train_target)]
for name, pred, target in sets:
if len(target) < 3:
print('There are less than 3 values to predict, skipping.', file=sys.stderr)
continue
print('%s set:' % name,
'R2_score: %.4f' % r2_score(target, pred),
'Rp: %.4f' % pearsonr(target, pred)[0],
'RMSE: %.4f' % rmse(target, pred),
'SD: %.4f' % standard_deviation_error(target, pred),
sep='\t', file=sys.stderr)
if sf_pickle is None:
return self.save('PLEC%s_p%i_l%i_pdbbind%i_s%i.pickle'
% (self.version, self.depth_protein,
self.depth_ligand, pdbbind_version, self.size))
else:
return self.save(sf_pickle)
def test_standard_deviation_error():
assert standard_deviation_error(values, good_values) < 1.1
assert standard_deviation_error(values, poor_values) > 2e4
def train(self, home_dir=None, sf_pickle=None, pdbbind_version=2016,
ignore_json=False):
if not home_dir:
home_dir = path_join(dirname(__file__), 'PLECscore')
desc_path = path_join(home_dir, 'plecscore_descs_p%i_l%i.csv.gz' %
(self.depth_protein, self.depth_ligand))
json_path = path_join(
home_dir, 'plecscore_%s_p%i_l%i_s%i_pdbbind%i.json' %
(self.version, self.depth_protein,
self.depth_ligand, self.size, pdbbind_version))
if (self.version in ['linear'] and # TODO: support other models
isfile(json_path) and
not ignore_json):
print('Loading pretrained PLECscore %s with depths P%i L%i on '
'PDBBind v%i'
% (self.version, self.depth_protein, self.depth_ligand,
pdbbind_version), file=sys.stderr)
with open(json_path) as json_f:
json_data = json.load(json_f)
for k, v in json_data.items():
if isinstance(v, list):
if isinstance(v[0], list):
v = [np.array(x) for x in v]
else:
v = np.array(v)
setattr(self.model, k, v)
else:
# blacklist core set 2013 and astex
pdbids_blacklist = [
'3ao4', '3i3b', '1uto', '1ps3', '1qi0', '3g2z', '3dxg', '3l7b',
'3mfv', '3b3s', '3kgp', '3fk1', '3fcq', '3lka', '3udh', '4gqq',
'3imc', '2xdl', '2ymd', '1lbk', '1bcu', '3zsx', '1f8d', '3muz',
'2v00', '1loq', '3n7a', '2r23', '3nq3', '2hb1', '2w66', '1n2v',
'3kwa', '3g2n', '4de2', '3ozt', '3b3w', '3cft', '3f3a', '2qmj',
'3f80', '1a30', '1w3k', '3ivg', '2jdy', '3u9q', '3pxf', '2wbg',
'1u33', '2x0y', '3mss', '1vso', '1q8t', '3acw', '3bpc', '3vd4',
'3cj2', '2brb', '1p1q', '2vo5', '3d4z', '2gss', '2yge', '3gy4',
'3zso', '3ov1', '1w4o', '1zea', '2zxd', '3ueu', '2qft', '1gpk',
'1f8b', '2jdm', '3su5', '2wca', '3n86', '2x97', '1n1m', '1o5b',
'2y5h', '3ehy', '4des', '3ebp', '1q8u', '4de1', '3huc', '3l4w',
'2vl4', '3coy', '3f3c', '1os0', '3owj', '3bkk', '1yc1', '1hnn',
'3vh9', '3bfu', '1w3l', '3k5v', '2qbr', '1lol', '10gs', '2j78',
'1r5y', '2weg', '3uo4', '3jvs', '2yfe', '1sln', '2iwx', '2jdu',
'4djv', '2xhm', '2xnb', '3s8o', '2zcr', '3oe5', '3gbb', '2d3u',
'3uex', '4dew', '1xd0', '1z95', '2vot', '1oyt', '2ole', '3gcs',
'1kel', '2vvn', '3kv2', '3pww', '3su2', '1f8c', '2xys', '3l4u',
'2xb8', '2d1o', '2zjw', '3f3e', '2g70', '2zwz', '1u1b', '4g8m',
'1o3f', '2x8z', '3cyx', '2cet', '3ag9', '2pq9', '3l3n', '1nvq',
'2cbj', '2v7a', '1h23', '2qbp', '3b68', '2xbv', '2fvd', '2vw5',
'3ejr', '3f17', '3nox', '1hfs', '1jyq', '2pcp', '3ge7', '2wtv',
'2zcq', '2obf', '3e93', '2p4y', '3dd0', '3nw9', '3uri', '3gnw',
'3su3', '2xy9', '1sqa', '3fv1', '2yki', '3g0w', '3pe2', '1e66',
'1igj', '4tmn', '2zx6', '3myg', '4gid', '3utu', '1lor', '1mq6',
'2x00', '2j62', '4djr', '1gm8', '1gpk', '1hnn', '1hp0', '1hq2',
'1hvy', '1hwi', '1hww', '1ia1', '1j3j', '1jd0', '1jje', '1ke5',
'1kzk', '1l2s', '1l7f', '1lpz', '1m2z', '1mmv', '1mzc', '1n1m',
'1n2v', '1n46', '1nav', '1of1', '1of6', '1opk', '1oq5', '1owe',
'1oyt', '1p2y', '1p62', '1pmn', '1q1g', '1q41', '1q4g', '1r1h',
'1r55', '1r58', '1r9o', '1s19', '1s3v', '1sg0', '1sj0', '1sq5',
'1sqn', '1t40', '1t46', '1t9b', '1tow', '1tt1', '1u1c', '1uml',
'1unl', '1uou', '1v0p', '1v48', '1v4s', '1vcj', '1w1p', '1w2g',
'1xm6', '1xoq', '1xoz', '1y6b', '1ygc', '1yqy', '1yv3', '1yvf',
'1ywr', '1z95', '2bm2', '2br1', '2bsm']
# use remote csv if it's not present
if not isfile(desc_path):
branch = 'master' # define branch/commit
desc_url = ('https://raw.githubusercontent.com/oddt/oddt/%s'
'/oddt/scoring/functions/PLECscore/'
'plecscore_descs_p%i_l%i.csv.gz' %
(branch, self.depth_protein, self.depth_ligand))
warnings.warn('The CSV for PLEC P%i L%i is missing. Trying to '
'get it from ODDT GitHub.' % (self.depth_protein,
self.depth_ligand))
# download and save CSV
pd.read_csv(desc_url, index_col='pdbid').to_csv(
desc_path, compression='gzip')
# set PLEC size to unfolded
super(PLECscore, self)._load_pdbbind_desc(
desc_path,
train_set=('general', 'refined'),
pdbbind_version=pdbbind_version,
train_blacklist=pdbids_blacklist,
fold_size=self.size,
)
print('Training PLECscore %s with depths P%i L%i on PDBBind v%i'
% (self.version, self.depth_protein, self.depth_ligand,
pdbbind_version), file=sys.stderr)
self.model.fit(self.train_descs, self.train_target)
sets = [
('Test', self.model.predict(self.test_descs), self.test_target),
('Train', self.model.predict(self.train_descs), self.train_target)]
if self.version == 'rf':
sets.append(('OOB', self.model.oob_prediction_, self.train_target))
for name, pred, target in sets:
print('%s set:' % name,
'R2_score: %.4f' % r2_score(target, pred),
'Rp: %.4f' % pearsonr(target, pred)[0],
'RMSE: %.4f' % rmse(target, pred),
'SD: %.4f' % standard_deviation_error(target, pred),
sep='\t', file=sys.stderr)
if sf_pickle is None:
return self.save('PLEC%s_p%i_l%i_pdbbind%i_s%i.pickle'
% (self.version, self.depth_protein,
self.depth_ligand, pdbbind_version, self.size))
else:
return self.save(sf_pickle)
def test_standard_deviation_error():
assert_less(standard_deviation_error(values, good_values), 1.1)
assert_greater(standard_deviation_error(values, poor_values), 5e4)