Question about more datasets and more baselines

[Dinxin]

Why don't you give the associated experimental results on some regression tasks in the MoleculeNet benchmark? Additionally, there are some pertaining techniques tailored for molecule data such as mol2vec, seq2seq fingerprint. However, I don't see the model comparison.

[weihua916]

Great question! In this paper, we focus on GNNs. Testing all those techniques is beyond the scope of a single paper, but it is definitely an important work to do.

[badtom]

Given that you do a "graph-level multi-task supervised pre-training" on the preprocessed ChEMBL dataset from the publication "Large-scale comparison of machine learning methods for drug target prediction on ChEMBL" did you perhaps also check whether your GNN outperforms the multi-task feedforward neural network from this publication on this set (note that the best AUC ROC results for a random split were reported by the authors when using additional data normalization in http://www.bioinf.jku.at/research/lsc/report.pdf)?