def test_fps_5(data):
fps = Fingerprints(featurizers='ECFP', n_jobs=1, input_type='any', on_errors='nan', n_bits=100)
ret = fps.transform(data['err_smis'])
assert isinstance(ret, pd.DataFrame)
assert ret.shape == (4, 100)
assert np.isnan(ret.values[1][10])
assert np.isnan(ret.values[2][20])
def test_fps_6(data):
fps = Fingerprints(n_jobs=1,
input_type='any',
on_errors='nan',
counting=True)
ret = fps.transform(data['smis'])
assert isinstance(ret, pd.DataFrame)
assert ret.shape == (4, 16751)
def test_fps_3(data):
fps = Fingerprints(n_jobs=1, input_type='any')
try:
fps.transform(data['mols'] + data['smis'])
except BaseException as e:
print(e)
assert False, 'should not got error'
else:
assert True
def test_fps_4(data):
fps = Fingerprints(n_jobs=1, input_type='any')
with pytest.raises(ValueError):
fps.transform(data['err_smis'])
fps = Fingerprints(n_jobs=1, input_type='any', on_errors='nan')
ret = fps.transform(data['err_smis'])
assert isinstance(ret, pd.DataFrame)
assert ret.shape == (4, 10607)
assert np.isnan(ret.values[1][10])
assert np.isnan(ret.values[2][20])
def test_fps_1(data):
fps = Fingerprints(n_jobs=1)
try:
fps.transform(data['mols'])
except:
assert False, 'should not got error'
else:
assert True
try:
fps.transform(data['smis'])
except TypeError:
assert True
else:
assert False, 'should not got error'
def main():
# load in-house data from csv file
data = pd.read_csv("./iQSPR_sample_data.csv")
data_ss = data.sample(3000).reset_index()
from xenonpy.descriptor import Fingerprints
RDKit_FPs = Fingerprints(featurizers=['ECFP'], input_type='smiles')
# make models
# prd_mdls = make_forward_model_with_iqspr_tools(data_ss, RDKit_FPs)
prd_mdls, mdls = make_forward_model(data_ss, RDKit_FPs)
# learn NGrams by method 0 (not recommended)
# n_gram = learn_n_gram0(data_ss['SMILES'])
# learn NGrams by method 1 (recommended)
# n_gram = learn_n_gram1(data_ss['SMILES'])
# learn NGrams by method2 (recommended)
n_gram = learn_n_gram2(data_ss['SMILES'])
n_gram.set_params(del_range=[1, 20], max_len=500, reorder_prob=0.5)
# set up initial molecules for iQSPR
np.random.seed(201903) # fix the random seed
cans = [Chem.MolToSmiles(Chem.MolFromSmiles(smi)) for i, smi in enumerate(data_ss['SMILES'])
if (data_ss['H**O-LUMO gap'].iloc[i] > 4)]
init_samples = np.random.choice(cans, 25)
# set up annealing schedule in iQSPR
beta = np.hstack(
[np.linspace(0.01, 0.2, 20), np.linspace(0.21, 0.4, 10), np.linspace(0.4, 1, 10), np.linspace(1, 1, 10)])
iqspr_results = run_iqspr(prd_mdls, n_gram, init_samples, beta)
visualize(iqspr_results, RDKit_FPs, mdls, data_ss, beta)
def test_fps_4(data):
fps = Fingerprints(n_jobs=1, input_type='any')
try:
fps.transform(data['err_smis'])
except ValueError:
assert True
else:
assert False, 'should not got error'
fps = ECFP(n_jobs=1, input_type='any', on_errors='nan')
try:
ret = fps.transform(data['err_smis'])
assert pd.DataFrame(data=ret).shape == (4, 2048)
assert np.isnan(ret[1][10])
assert np.isnan(ret[2][20])
except:
assert False
else:
assert True
def property_plot(smile, plot_path):
P_fp = Fingerprints(featurizers=fp_type,
input_type='smiles',
on_errors='nan').transform([smile])
P_fp = P_fp.dropna()
PolyGeno_property = pd.read_csv(
"./data/forward_train_Polymer_genome_new_asterisk_R.csv")
DC = np.ravel(PolyGeno_property['Dielectric.Constant'])
GTT = np.ravel(PolyGeno_property['Glass.Transition.Temperature'])
singleFP2dc = pickle.load(open('./models/P_SMILES2DC_ElasticNet.sav',
'rb'))
singleFP2gtt = pickle.load(
open("./models/P_SMILES2GTT_ElasticNet.sav", 'rb'))
property_region = {
'Dielectric_Constant': [3, 4],
'Glass_Transition_Temperature': [300, 350]
}
property_minmax = {
'Dielectric_Constant': [min(DC), max(DC)],
'Glass_Transition_Temperature': [min(GTT), max(GTT)]
}
plt.figure(figsize=(10, 10))
plt.xlim(property_minmax["Dielectric_Constant"][0] - 1,
property_minmax["Dielectric_Constant"][1] + 1)
plt.ylim(property_minmax["Glass_Transition_Temperature"][0] - 20,
property_minmax["Glass_Transition_Temperature"][1] + 20)
plt.plot(DC, GTT, 'k.', markersize=5, alpha=0.4, label='existing data')
if len(P_fp) == 1:
dc_pre = singleFP2dc.predict(P_fp)
gtt_pre = singleFP2gtt.predict(P_fp)
plt.scatter(dc_pre,
gtt_pre,
s=250,
c='g',
edgecolor='k',
label='product')
plt.xlabel("Dielectric Constant", fontsize=20)
plt.ylabel("Glass Transition Temperature ($^\circ$C)", fontsize=20)
plt.title("Properties of chemical structures", fontsize=20)
plt.legend(loc='lower right', fontsize=20)
plt.xticks(fontsize=20)
plt.yticks(fontsize=20)
plt.plot([
property_region["Dielectric_Constant"][0],
property_region["Dielectric_Constant"][0],
property_region["Dielectric_Constant"][1],
property_region["Dielectric_Constant"][1],
property_region["Dielectric_Constant"][0]
], [
property_region["Glass_Transition_Temperature"][0],
property_region["Glass_Transition_Temperature"][1],
property_region["Glass_Transition_Temperature"][1],
property_region["Glass_Transition_Temperature"][0],
property_region["Glass_Transition_Temperature"][0]
], 'r--')
plt.savefig(plot_path)
plt.close()
if len(P_fp) == 0:
dc_pre = 0
gtt_pre = 0
dc_pre = round(dc_pre[0], 2)
gtt_pre = round(gtt_pre[0], 2)
return (dc_pre, gtt_pre)
def test_fps_3(data):
fps = Fingerprints(n_jobs=1, input_type='any')
fps.transform(data['mols'] + data['smis'])
def test_fps_2(data):
fps = Fingerprints(n_jobs=1, input_type='smiles')
with pytest.raises(TypeError):
fps.transform(data['mols'])
fps.transform(data['smis'])
def test_fps_1(data):
fps = Fingerprints(n_jobs=1)
fps.transform(data['mols'])
with pytest.raises(TypeError):
fps.transform(data['smis'])