def create_feedforwardnet(civ_dim, hparams, compound_model, protein_profile):
if hparams["prot"]["model_type"] == "pcnna":
p = 2 * hparams["prot"]["dim"]
base_model = ProtCnnForward(*create_prot_net(hparams, protein_profile),
nn.Sequential(compound_model, nn.Linear(*civ_dim)))
else:
base_model = PairSequential(mod1=(compound_model,), mod2=(create_prot_net(hparams, protein_profile),))
p = np.sum(civ_dim)
fcn_args = []
fcn_layers = hparams["hdims"]
if not isinstance(fcn_layers, list):
fcn_layers = [fcn_layers]
for dim in fcn_layers:
conf = FcnArgs(in_features=p,
out_features=dim,
activation='relu',
batch_norm=True,
dropout=hparams["dprob"])
fcn_args.append(conf)
p = dim
fcn_args.append(FcnArgs(in_features=p, out_features=hparams["output_dim"]))
fcn_layers = create_fcn_layers(fcn_args)
model = nn.Sequential(base_model, *fcn_layers)
return model
def create_ecfp_net(hparams):
fcn_args = []
for i in range(len(hparams["ecfp"]["ecfp_hdims"]) - 1):
conf = FcnArgs(in_features=hparams["ecfp"]["ecfp_hdims"][i],
out_features=hparams["ecfp"]["ecfp_hdims"][i + 1],
activation='relu',
batch_norm=True,
dropout=hparams["dprob"])
fcn_args.append(conf)
fcn_args.append(FcnArgs(in_features=hparams["ecfp"]["ecfp_hdims"][-1], out_features=hparams["latent_dim"]))
fcn_args.append(FcnArgs(in_features=hparams["latent_dim"], out_features=1))
model = nn.Sequential(*create_fcn_layers(fcn_args))
return model
def create_discriminator_net(hparams):
fcn_args = []
p = hparams["neigh_dist"]
layers = hparams["disc_hdims"]
if not isinstance(layers, list):
layers = [layers]
for dim in layers:
conf = FcnArgs(in_features=p,
out_features=dim,
activation='relu',
batch_norm=True,
dropout=hparams["dprob"])
fcn_args.append(conf)
p = dim
fcn_args.append(
FcnArgs(in_features=p,
out_features=hparams['output_dim'],
activation="sigmoid"))
layers = create_fcn_layers(fcn_args)
model = nn.Sequential(*layers)
return model
def create_integrated_net(hparams):
# segment 1 - graphconv
gconv_model = GraphConvSequential(
GraphConvLayer(in_dim=75, out_dim=64), nn.BatchNorm1d(64), nn.ReLU(),
GraphPool(), GraphConvLayer(in_dim=64, out_dim=64), nn.BatchNorm1d(64),
nn.ReLU(), GraphPool(),
nn.Linear(in_features=64, out_features=hparams["gconv_dim"]),
nn.BatchNorm1d(hparams["gconv_dim"]), nn.ReLU(),
nn.Dropout(hparams["dprob"]), GraphGather())
# segment 2 - fingerprint
fp_net = nn.Identity()
# segment 3 - protein
prot_net = nn.Identity()
civ_net = PairSequential(
(PairSequential(mod1=(gconv_model, ), mod2=(fp_net, )), ),
(prot_net, ))
civ_dim = hparams["prot_dim"] + hparams["gconv_dim"] * 2 + hparams["fp_dim"]
fcn_args = []
p = civ_dim
layers = hparams["hdims"]
if not isinstance(layers, list):
layers = [layers]
for dim in layers:
conf = FcnArgs(in_features=p,
out_features=dim,
activation='relu',
batch_norm=True,
dropout=hparams["dprob"])
fcn_args.append(conf)
p = dim
fcn_args.append(FcnArgs(in_features=p, out_features=hparams['output_dim']))
fcn_layers = create_fcn_layers(fcn_args)
model = nn.Sequential(civ_net, *fcn_layers)
return model