Install

import pandas as pd
from rdkit import Chem

# URL to the Delaney dataset in CSV format
url = "https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/delaney-processed.csv"

# Download and load the dataset
df = pd.read_csv(url)

selected_df = df[['smiles', 'measured log solubility in mols per litre']]

# Renaming the columns for clarity
selected_df.columns = ['SMILES', 'Solubility']

# Save the selected data to a new CSV file
selected_df.to_csv('solubility_dataset.csv', index=False)

import pandas as pd
from gnn import smiles_to_graph, SimpleGNN
from torch import optim, nn
from numpy import mean
from torch_geometric.data import DataLoader

df = pd.read_csv("solubility_dataset.csv")

model = SimpleGNN()
optimizer = optim.Adam(model.parameters(), lr=1e-3)
criterion = nn.MSELoss()

graphs = [smiles_to_graph(smiles, float(sol)) for smiles, sol in zip(df['SMILES'], df['Solubility'].values)]

loader = DataLoader(graphs, batch_size=32, shuffle=True)

loss_l = []
for step in range(1):
    for data in loader:
        optimizer.zero_grad()
        prediction = model(data)
        loss = criterion(prediction, data.target)
        loss.backward()
        optimizer.step()
        loss_l.append(loss.item())
    print(mean(loss_l))