def main(argv):
del argv # unused.
config_name = '/data/junyoung/workspace/Mol_DQN/models/multi_logp_qed_model/config_2'
all_cid = '/data/junyoung/workspace/Mol_DQN/Config/all_cid'
with open(config_name) as f:
hparams = json.load(f)
with open(all_cid) as f:
all_mols = json.load(f)
# init_mol = ["CNC(=O)/C(C#N)=C(/[O-])C1=NN(c2cc(C)ccc2C)C(=O)CC1"]
environment = Multi_LogP_QED_Molecule(hparams=hparams,
molecules=all_mols)
dqn = deep_q_networks.DeepQNetwork(
hparams=hparams,
q_fn=functools.partial(
deep_q_networks.Q_fn_neuralnet_model,
hparams=hparams))
Trainer =trainer.Trainer(
hparams=hparams,
environment=environment,
model=dqn)
Trainer.run_training()
config.write_hparams(hparams, os.path.join(hparams['save_param']['model_path'], 'config.json'))
def main(argv):
del argv # unused.
config_name = '/home/junyoung/workspace/Mol_DQN/models/logp_model/config'
all_cid = '/home/junyoung/workspace/Mol_DQN/Config/all_cid'
with open(config_name) as f:
hparams = json.load(f)
# with open(all_cid) as f:
# all_mols = json.load(f)
environment = LogP_Molecule(hparams=hparams,
molecules=None)
dqn = deep_q_networks.DeepQNetwork(
hparams=hparams,
q_fn=functools.partial(
deep_q_networks.Q_fn_neuralnet_model, hparams=hparams))
Trainer =trainer.Trainer(
hparams=hparams,
environment=environment,
model=dqn)
Trainer.run_training()
config.write_hparams(hparams, os.path.join(hparams['save_param']['model_dir'], 'config.json'))
def main(argv):
del argv # unused.
config_name = '/home/junyoung/workspace/Mol_DQN_ADMET_v2/models/admet_qed_model/config_1'
# all_cid = '/home/junyoung/workspace/Mol_DQN_ADMET/Config/all_cid'
logS_model_path = '/home/junyoung/workspace/ADMET/ADMETlab/regression_model/logS/logS_Model_v3.pkl'
caco_model_path = '/home/junyoung/workspace/ADMET/ADMETlab/regression_model/caco2/caco2_Model_v3.pkl'
cyp3a4_model_path = '/home/junyoung/workspace/ADMET/ADMETlab/classification_model/CYP3A4-inhibitor/CYP_inhibitor_3A4_SVC_ecfp4_model_v3.pkl'
# ppb_model_path = '/home/junyoung/workspace/ADMET/ADMETlab/regression_model/PPB/PPB_Model_v3.pkl'
t_model_path = '/home/junyoung/workspace/ADMET/ADMETlab/regression_model/T/T_Model_v3.pkl'
ld50_model_path = '/home/junyoung/workspace/ADMET/ADMETlab/regression_model/LD50/LD50_Model_v3.pkl'
with open(config_name) as f:
hparams = json.load(f)
# with open(all_cid) as f:
# all_mols = json.load(f)
with open(logS_model_path, 'rb') as f:
logS_model = pickle.load(f, encoding="latin1")
# with open(ppb_model_path, 'rb') as f:
# ppb_model = pickle.load(f, encoding="latin1")
with open(caco_model_path, 'rb') as f:
caco_model = pickle.load(f, encoding="latin1")
with open(cyp3a4_model_path, 'rb') as f:
cyp3a4_model = pickle.load(f, encoding="latin1")
with open(t_model_path, 'rb') as f:
t_model = pickle.load(f, encoding="latin1")
with open(ld50_model_path, 'rb') as f:
ld50_model = pickle.load(f, encoding="latin1")
##To calculate SA score##
SA_model = molecules_rules.readSAModel()
all_mols = ["C1(=CC(=CC=C1NC2=NC(=C(C=N2)Cl)CC3=CC=CC=C3C)[P](C)(C)=O)O"]
mol = Chem.MolFromSmiles(all_mols[0])
rings = mol.GetRingInfo().BondRings()
ring_scaffolds = [
Chem.MolToSmiles(Chem.PathToSubmol(mol, ring)) for ring in rings
]
environment = ADMET_QED_Molecule(hparams=hparams,
molecules=all_mols,
SA_model=SA_model,
logS_model=logS_model,
caco_model=caco_model,
cyp3a4_model=cyp3a4_model,
t_model=t_model,
ld50_model=ld50_model,
scaffold=ring_scaffolds,
record_path=True)
dqn = deep_q_networks.DeepQNetwork(
hparams=hparams,
q_fn=functools.partial(deep_q_networks.Q_fn_neuralnet_model,
hparams=hparams))
Trainer = trainer.Trainer(hparams=hparams,
environment=environment,
model=dqn)
Trainer.run_training()
log = logging.getLogger("rhythm")
if args.module is None:
raise ValueError("Provide a module argument (see --help)")
configure_logging(args.module, args.verbose)
# seed for reproducability
torch.manual_seed(args.seed)
np.random.seed(args.seed)
if hasattr(args, "device") and args.device != "cpu":
torch.cuda.manual_seed_all(args.seed)
log.info("Arguments: {}".format(args))
if args.module == "process":
data_process = process.Process(args)
data_process.process()
elif args.module == "compute-mean":
mean_compute = ComputeMean(args)
mean_compute.compute()
elif args.module == "train":
trainer = trainer.Trainer(args)
if args.test:
trainer.load()
trainer.test()
else:
trainer.train()
else:
raise ValueError("Unknown module: {}".format(args.module))