RetroGNN: Approximating Retrosynthesis by Graph Neural Networks for De Novo Drug Design
This work provides a faster and more accurate method for filtering synthesizable molecules, which is crucial for researchers in de novo drug design to generate more realistic and effective drug candidates.
This paper addresses the issue of generating chemically unfeasible molecules in de novo drug design by training graph neural networks (GNNs) to approximate retrosynthesis planning software. The GNN-biased search finds molecules predicted to be more effective antibiotics than those from the ZINC database, while maintaining drug-like properties and synthesizability. The GNN achieves a 10^5 times speed-up in filtering hard-to-synthesize molecules compared to traditional retrosynthesis software.
De novo molecule generation often results in chemically unfeasible molecules. A natural idea to mitigate this problem is to bias the search process towards more easily synthesizable molecules using a proxy for synthetic accessibility. However, using currently available proxies still results in highly unrealistic compounds. We investigate the feasibility of training deep graph neural networks to approximate the outputs of a retrosynthesis planning software, and their use to bias the search process. We evaluate our method on a benchmark involving searching for drug-like molecules with antibiotic properties. Compared to enumerating over five million existing molecules from the ZINC database, our approach finds molecules predicted to be more likely to be antibiotics while maintaining good drug-like properties and being easily synthesizable. Importantly, our deep neural network can successfully filter out hard to synthesize molecules while achieving a $10^5$ times speed-up over using the retrosynthesis planning software.