def main(epoch):
decodings2 = read_decodings()
in_mols = np.load("History/in-{}.npy".format(epoch))
out_mols = np.load("History/out-{}.npy".format(epoch))
in_mols = [decode(m, decodings2) for m in in_mols]
out_mols = [safe_decode(m, decodings2) for m in out_mols]
use = [(not out_mols[i] is None) and \
Chem.MolToSmiles(out_mols[i]) != Chem.MolToSmiles(in_mols[i])
for i in range(len(out_mols))]
plot_mols = [[m1, m2] for m1, m2, u in zip(in_mols, out_mols, use) if u]
order = [
np.sum(evaluate_chem_mol(out_mols[i])) for i in range(len(out_mols))
if use[i]
]
plot_mols = [x for _, x in sorted(zip(order, plot_mols), reverse=True)]
plot_mols = [x for y in plot_mols for x in y]
# #
plot = Draw.MolsToGridImage(plot_mols[:50], molsPerRow=2)
plot.show()
plot.save("out_grid.png")
def safe_decode(x, decodings):
try:
m = decode(x, decodings)
Chem.Kekulize(m)
return m
except:
return None
def main(epoch, savefile=None, imagefile=None):
decodings2 = read_decodings()
in_mols = np.load("History/in-{}.npy".format(epoch))
out_mols = np.load("History/out-{}.npy".format(epoch))
in_mols = [decode(m, decodings2) for m in in_mols]
out_mols = [safe_decode(m, decodings2) for m in out_mols]
use = [(not out_mols[i] is None) and \
Chem.MolToSmiles(out_mols[i]) != Chem.MolToSmiles(in_mols[i])
for i in range(len(out_mols))]
plot_mols = [[m1, m2] for m1, m2, u in zip(in_mols, out_mols, use) if u]
# order = [np.sum(evaluate_chem_mol(out_mols[i])) for i in range(len(out_mols)) if use[i]]
# plot_mols = [x for _,x in sorted(zip(order,plot_mols),key=lambda x:x[0],
# reverse=True)]
plot_mols = [x for y in plot_mols for x in y]
plot = Draw.MolsToGridImage(plot_mols[:50], molsPerRow=2)
if not imagefile is None:
plot.save(imagefile)
plot.show()
if not savefile is None:
with open(savefile, "w") as f:
f.write("Initial molecule ; Modified molecule\n")
for i in range(0, len(plot_mols), 2):
f.write(
f'{Chem.MolToSmiles(plot_mols[i])} ; {Chem.MolToSmiles(plot_mols[i+1])}\n'
)
def bunch_eval(fs, epoch, decodings):
global evaluated_mols
keys = []
od = OrderedDict()
#total_molecules = len(fs)
#print("Evaluating totally {} molecules".format(total_molecules))
for f in fs:
key = get_key(f)
keys.append(key)
od[key] = ([False,False])
#print(len(od))
to_evaluate = []
i = 0
unused = keys.copy()
for key in keys:
if key in evaluated_mols:
od[key] = evaluated_mols[key][0]
while key in unused:
unused.remove(key)
else:
try:
mol = decode(fs[i], decodings)
od[key] = len(to_evaluate)
to_evaluate.append(mol)
# evaluated_mols[key] = (np.array(ret_val), epoch)
except:
od[key] = [False,-10]
evaluated_mols[key] = (np.asarray([False,-10]),epoch)
i = i + 1
print('New molecules for evaluation: {}'.format(len(to_evaluate)))
if len(to_evaluate)!=0:
Evaluations = bunch_evaluation(to_evaluate)
#print("Length of Evaluations {}".format(len(Evaluations)))
assert len(Evaluations) == len(to_evaluate)
for i in range(len(Evaluations)):
for key in unused:
if od[key] == i:
value = Evaluations[i]
od[key] = value
evaluated_mols[key] = (np.array(value),epoch)
ret_vals = []
with open('./ret_vals.pkl','wb') as f:
pickle.dump(ret_vals,f)
for key in keys:
ret_vals.append(np.asarray(od[key]))
ret_vals = np.asarray(ret_vals)
print('Shape of return values {}'.format(ret_vals.shape))
return (ret_vals)
# =============================================================================
# df = pd.read_csv('./out.csv',engine="python")
# for i in range(len(df)):
# print("Molecule number {}".format(i+1))
# mol1 = Chem.MolFromSmiles(df.iloc[i,0])
# mol2 = Chem.MolFromSmiles(df.iloc[i,1])
# print(get_pIC(mol1),get_pIC(mol2))
# =============================================================================
def evaluate_mol(fs, epoch, decodings):
global evaluated_mols
key = get_key(fs)
if evaluated_mols.has_key(key):
return evaluated_mols[key][0]
try:
mol = decode(fs, decodings)
ret_val = evaluate_chem_mol(mol)
except:
ret_val = [False] * 4
evaluated_mols[key] = (np.array(ret_val), epoch)
return np.array(ret_val)