The dataset is quite similar to the MUTAG introduced in the practical session, but with more data. Take care to first check the dimensionality of your node features to adapt your GNN.