def getGraphX(AID):
'''Takes in AID, finds graphreps for pytroch implementation'''
AID_path = os.path.join('/home/gabriel/Dropbox/UCL/Thesis/Data/', AID)
save_path = AID_path+ '/' + AID +'mol_processed.pkl'
pickle_off = open(save_path,'rb')
activity_table=pickle.load(pickle_off)
pickle_off.close()
graph_rep_list = [mol2graph.mol2vec(m) for m in activity_table['MOL']]
AID_and_graph_rep = pd.DataFrame()
AID_and_graph_rep['Graph Rep']= graph_rep_list
AID_and_graph_rep['PUBCHEM_CID'] = activity_table['PUBCHEM_CID']
main_aid_save_path = AID_path+ '/' + AID +'_processed.pkl'
pickle_off = open(main_aid_save_path,'rb')
main_activity_table=pickle.load(pickle_off)
pickle_off.close()
main_activity_table = main_activity_table.merge(AID_and_graph_rep,on='PUBCHEM_CID')
save_df = True
new_aid_save_path = main_aid_save_path = AID_path+ '/' + AID +'graph_processed.pkl'
if save_df == True:
main_activity_table.to_pickle(new_aid_save_path)
return main_activity_table
#%%
'''Classifier Section'''
'''SVM'''
test_size=0.2,
random_state=2562)
labels = np.array([
1 if x == 'Active' else 0
for x in activity_table['PUBCHEM_ACTIVITY_OUTCOME']
])
for big_train_ind, big_test_ind in big_splitter.split(
activity_table, activity_table['PUBCHEM_ACTIVITY_OUTCOME']):
train_X = np.atleast_2d(activity_table['MOL'].iloc[big_train_ind]).T
train_y = labels[big_train_ind]
test_X = np.atleast_2d(activity_table['MOL'].iloc[big_test_ind]).T
test_y = labels[big_test_ind]
train_X_oversampled, train_y_oversampled = ros.fit_resample(
train_X, train_y)
train_X = [
mol2graph.mol2vec(m) for m in np.squeeze(train_X_oversampled)
]
test_X = [mol2graph.mol2vec(m) for m in np.squeeze(test_X)]
#attach train labels to data
for data, label in zip(train_X, train_y_oversampled):
data.y = torch.tensor([[label]], dtype=torch.float)
for data, label in zip(test_X, test_y):
data.y = torch.tensor([[label]], dtype=torch.float)
train_loader = DataLoader(train_X,
batch_size=128,
shuffle=True,
drop_last=False,
num_workers=8)
test_loader = DataLoader(test_X,
batch_size=128,
shuffle=True,
f'Epoch: {epoch}, Loss: {epoch_loss:.3f}, Train acc: {train_acc:.3f}, Val acc: {val_acc:.3f}',
)
return hist
if __name__ == '__main__':
get_data()
train_mols = [m for m in Chem.SDMolSupplier('solubility.train.sdf')]
test_mols = [m for m in Chem.SDMolSupplier('solubility.test.sdf')]
sol_cls_dict = {'(A) low': 0, '(B) medium': 1, '(C) high': 2}
print(sol_cls_dict)
train_x = [mol2graph.mol2vec(m) for m in train_mols]
for i, data in enumerate(train_x):
y = sol_cls_dict[train_mols[i].GetProp('SOL_classification')]
data.y = torch.tensor([y], dtype=torch.long)
test_x = [mol2graph.mol2vec(m) for m in test_mols]
for i, data in enumerate(test_x):
y = sol_cls_dict[test_mols[i].GetProp('SOL_classification')]
data.y = torch.tensor([y], dtype=torch.long)
print(f'Number of graphs: {len(train_x)}')
first_sample = train_x[0]
print(f'Looking at first example ..')
print(f'\t{train_x[0]}')
print(f'\t # of nodes: {first_sample.x.shape[0]}')
print(f'\t # of features per node: {first_sample.x.shape[1]}')