Continuous Representation of Molecules Using Graph Variational Autoencoder
This work addresses molecular representation for drug design and property prediction, presenting an incremental improvement over existing methods.
The authors tackled the problem of continuous molecular representation by proposing a graph variational autoencoder that operates on 2D molecular graphs, using a side predictor to prune the latent space and improve adjacency tensor generation. They demonstrated superior performance compared to SMILES-based RNN methods, though no specific numerical results were provided.
In order to continuously represent molecules, we propose a generative model in the form of a VAE which is operating on the 2D-graph structure of molecules. A side predictor is employed to prune the latent space and help the decoder in generating meaningful adjacency tensor of molecules. Other than the potential applicability in drug design and property prediction, we show the superior performance of this technique in comparison to other similar methods based on the SMILES representation of the molecules with RNN based encoder and decoder.