dim_edge = DE.shape[3]
dim_y = DY.shape[1]
atom_list = ['B', 'C', 'N', 'O', 'F', 'Si', 'P', 'S', 'Cl', 'Se', 'Br', 'I']
bond_list = ['SINGLE', 'DOUBLE', 'TRIPLE']
bpatt_dim = [3, 3, 2, 1, 2, 1]
print(':: preprocess data')
scaler = StandardScaler()
scaler.fit(DY)
DY = scaler.transform(DY)
mu_prior = np.mean(DY, 0)
cov_prior = np.cov(DY.T)
model = Model(n_node, dim_node, dim_edge, dim_y, atom_list, bpatt_dim,
mu_prior, cov_prior)
print(':: generate molecular graphs')
with model.sess:
model.saver.restore(model.sess, save_path)
total_count, valid_count, novel_count, unique_count, genmols = model.test(
10000, 0, Dsmi)
valid = valid_count / total_count
unique = unique_count / valid_count
novel = novel_count / valid_count
print(':: Valid:', valid * 100, 'Unique:', unique * 100, 'Novel:', novel * 100,
'GMean:', 100 * (valid * unique * novel)**(1 / 3))
print(':: preprocess data')
scaler = StandardScaler()
scaler.fit(DY)
DY = scaler.transform(DY)
mu_prior=np.mean(DY,0)
cov_prior=np.cov(DY.T)
model = Model(n_node, dim_node, dim_edge, dim_y, mu_prior, cov_prior)
print(':: generate molecular graphs')
with model.sess:
model.saver.restore(model.sess, save_path)
# unconditional generation
total_count, valid_count, novel_count, unique_count, genmols = model.test(10000, 0, Dsmi, atom_list)
valid=valid_count/total_count
unique=unique_count/valid_count
novel=novel_count/valid_count
list_Y=[]
for m in genmols:
mol = Chem.MolFromSmiles(m)
if dim_edge == 3: Chem.Kekulize(mol)
list_Y.append([Descriptors.ExactMolWt(mol), Descriptors.MolLogP(mol)])
print(':: unconditional generation results', len(genmols), np.mean(list_Y,0), np.std(list_Y,0))
print(':: Valid:', valid*100, 'Unique:', unique*100, 'Novel:', novel*100, 'GMean:', 100*(valid*unique*novel)**(1/3))
# conditional generation
print(':: preprocess data')
scaler = StandardScaler()
scaler.fit(DY)
DY = scaler.transform(DY)
mu_prior = np.mean(DY, 0)
cov_prior = np.cov(DY.T)
model = Model(n_node, dim_node, dim_edge, dim_y, mu_prior, cov_prior)
np.set_printoptions(precision=3, suppress=True)
with model.sess:
model.saver.restore(model.sess, save_path)
# unconditional generation
total_count, genmols = model.test(10000, 0, Dsmi, atom_list)
#print (total_count)
#valid=valid_count/total_count
#unique=unique_count/valid_count
#novel=novel_count/valid_count
#print(':: Valid:', valid*100, 'Unique:', unique*100, 'Novel:', novel*100, 'GMean:', 100*(valid*unique*novel)**(1/3))
list_Y = []
for mol in genmols:
#if dim_edge == 3: Chem.Kekulize(mol)
list_Y.append([isProperty(Chem.GetAdjacencyMatrix(mol))])
print(':: unconditional generation results', len(genmols),