def test_phagraphNN_generateFromSmiles_file():
import phagraphnn.utilities as ut
path = "./tests/data/twoD.smi"
mol = ut.CDPLmolFromSmiles(path,False)
assert(len(mol.atoms) == 22) #2D
mol_conf = ut.CDPLmolFromSmiles(path,True)
assert(len(mol_conf.atoms) == 37) #3D
def test_phagraphNN_generateFromSmiles_smile():
import phagraphnn.utilities as ut
data = ut.readChemblXls("./tests/data/CHE_3.xls")
mol =ut.CDPLmolFromSmiles(data[2][1],False) # same smiles then twoD.smi
mol_conf =ut.CDPLmolFromSmiles(data[2][1],True) # same smiles then twoD.smi
assert(len(mol.atoms) == 22) #2D
assert(len(mol_conf.atoms) == 37) #3D with Hydrogens
def test_PhaGat2_classification():
import phagraphnn.utilities as ut
from phagraphnn.PhaGraph import PhaGraph,PhaNode
data = ut.readChemblXls("./tests/data/CHE_3.xls")
graph_list = []
for i in range(0,len(data)):
graph = PhaGraph()
mol = ut.CDPLmolFromSmiles(data[i][1],True)
graph(ut.CDPLphaGenerator(None,mol,"lig_only"))
graph.setProperty("ic50",(0,1))
graph_list.append(graph)
from phagraphnn.DataPreparer import DataPreparer
loader = DataPreparer(graph_list,3,property_string="ic50",mpn="gru",is_path=False)
from phagraphnn.PhaGatModel2 import PhaGatModel2 as gat
import tensorflow as tf
seq = tf.keras.Sequential([
tf.keras.layers.Dense(16, activation='relu', input_shape=(128,),name="first_layer"),
tf.keras.layers.Dense(8, activation='relu',name="second_layer"),
tf.keras.layers.Dense(2,activation= None)],name="output_NN")
gat = gat(hidden_dim=32,output_nn=seq,regression = False)
lr = 0.001
gat.compile(loss=tf.keras.losses.BinaryCrossentropy(),
optimizer=tf.keras.optimizers.Adam(lr))
rec = tf.keras.metrics.MeanAbsoluteError()
for batch in loader:
inputs,af,other = batch
gat(inputs)
for epoch in range(0,10):
pred,loss = gat.train(inputs=inputs,outputs=af,learning_rate=lr)
def test_PhaGru():
import phagraphnn.utilities as ut
from phagraphnn.PhaGraph import PhaGraph,PhaNode
data = ut.readChemblXls("./tests/data/CHE_3.xls")
graph_list = []
for i in range(0,len(data)):
graph = PhaGraph()
mol = ut.CDPLmolFromSmiles(data[i][1],True)
graph(ut.CDPLphaGenerator(None,mol,"lig_only"))
graph.setProperty("ic50",data[i][2])
graph_list.append(graph)
from phagraphnn.DataPreparer import DataPreparer
loader = DataPreparer(graph_list,3,property_string="ic50",mpn="gru")
from phagraphnn.PhaGruMPN import PhaGruMPN as gru
from phagraphnn.PhaGruMPN2 import PhaGruMPN2 as gru2
from phagraphnn.PhaGruMPN3 import PhaGruMPN3 as gru3
import tensorflow as tf
seq = tf.keras.Sequential([
tf.keras.layers.Dense(16, activation='relu', input_shape=(32,),name="first_layer"),
tf.keras.layers.Dense(8, activation='relu',name="second_layer"),
tf.keras.layers.Dense(1,activation= None)],name="output_NN")
seq2 = tf.keras.Sequential([
tf.keras.layers.Dense(16, activation='relu', input_shape=(32,),name="first_layer"),
tf.keras.layers.Dense(8, activation='relu',name="second_layer"),
tf.keras.layers.Dense(1,activation= None)],name="output_NN")
seq3 = tf.keras.Sequential([
tf.keras.layers.Dense(16, activation='relu', input_shape=(32,),name="first_layer"),
tf.keras.layers.Dense(8, activation='relu',name="second_layer"),
tf.keras.layers.Dense(1,activation= None)],name="output_NN")
gru = gru(32,3,seq)
gru2 = gru2(32,3,seq2)
gru3 = gru3(32,3,seq3)
lr = 0.001
gru.compile(loss=tf.keras.losses.mse,
optimizer=tf.keras.optimizers.Adam(lr))
gru2.compile(loss=tf.keras.losses.mse,
optimizer=tf.keras.optimizers.Adam(lr))
gru3.compile(loss=tf.keras.losses.mse,
optimizer=tf.keras.optimizers.Adam(lr))
rec = tf.keras.metrics.MeanAbsoluteError()
for batch in loader:
inputs,af,other = batch
gru(inputs)
gru2(inputs)
gru3(inputs)
for epoch in range(0,10):
pred,loss = gru.train(inputs=inputs,outputs=af,learning_rate=lr)
pred,loss = gru2.train(inputs=inputs,outputs=af,learning_rate=lr)
pred,loss = gru3.train(inputs=inputs,outputs=af,learning_rate=lr)
import phagraphnn.utilities as ut
from phagraphnn.PhaGraph import PhaGraph, PhaNode
from phagraphnn.DataPreperer import DataPreparer
from phagraphnn.PhaGatModel import PhaGatModel as gat
from phagraphnn.PhaGatModel2 import PhaGatModel2 as gat2
from phagraphnn.PhaGruMPN import PhaGruMPN as gru
from phagraphnn.PhaGruMPN2 import PhaGruMPN2 as gru2
from phagraphnn.PhaGruMPN3 import PhaGruMPN3 as gru3
#%%
### load data
data = ut.readChemblXls("./tests/data/CHE_3.xls")
graph_list = []
for i in range(0, len(data)):
graph = PhaGraph()
mol = ut.CDPLmolFromSmiles(data[i][1], True)
graph(ut.CDPLphaGenerator(None, mol, "lig_only"))
graph.setProperty("ic50", data[i][2])
graph_list.append(graph)
loader = DataPreparer(graph_list,
3,
property_string="ic50",
mpn="gat",
is_path=False)
#%%
###generate network
# this is the MLP part and is put after the "encoding" of the graph
seq_gat = tf.keras.Sequential([
tf.keras.layers.Dense(