Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Junction Tree Variational Autoencoder for Molecular Graph Generation

About

We seek to automate the design of molecules based on specific chemical properties. In computational terms, this task involves continuous embedding and generation of molecular graphs. Our primary contribution is the direct realization of molecular graphs, a task previously approached by generating linear SMILES strings instead of graphs. Our junction tree variational autoencoder generates molecular graphs in two phases, by first generating a tree-structured scaffold over chemical substructures, and then combining them into a molecule with a graph message passing network. This approach allows us to incrementally expand molecules while maintaining chemical validity at every step. We evaluate our model on multiple tasks ranging from molecular generation to optimization. Across these tasks, our model outperforms previous state-of-the-art baselines by a significant margin.

Wengong Jin, Regina Barzilay, Tommi Jaakkola• 2018

Related benchmarks

TaskDatasetResultRank
Unconditional molecular generationMOSES
Validity100
39
Property optimizationZINC250k (test)
1st Order Metric0.925
33
Molecular Generationfa7
Top-Hit 5% Docking Score (kcal/mol)-7.683
29
Molecular Generation5ht1b
Docking Score (Top-Hit 5%, kcal/mol)-9.382
29
Molecular Generationjak2
Top-Hit 5% Docking Score (kcal/mol)-8.885
29
Molecular Generationparp1
Top-Hit 5% Docking Score (kcal/mol)-9.482
29
Molecular Generationbraf
Top-Hit 5% Docking Score (kcal/mol)-9.079
28
Constrained Property OptimizationZINC250K
Improvement1.68
27
Molecular GenerationZINC250k (test)
Validity100
26
Molecular Dockingjak2
Mean Docking Score-8.885
18
Showing 10 of 46 rows

Other info

Follow for update