Design of Synthesizable PROTACs through Synthesis Constrained Generative Model and Reinforcement Learning.

Proteolysis targeting chimeras (PROTACs) have emerged as a promising technology in degrading disease-related proteins for drug design. Recent deep generative models can accelerate PROTAC design, but the generated molecules are often difficult to synthesize. Here, we develop the SynPROTAC model, which integrates chemical reaction path-driven molecule assembly with reinforcement learning for the design of synthesizable PROTACs together with favorable binding properties. Specifically, the synthesis-constrained generative model employs a Graph Transformer-encoded warhead or E3 ligand as input and autoregressively samples reaction templates and building blocks through transformer-based decoder for PROTAC construction. The comprehensive evaluations indicated that SynPROTAC is capable of generating new PROTACs with feasible synthetic routes and reasonable physicochemical and binding-related properties. We further applied SynPROTAC to design PROTAC molecules degrading bromodomain-containing protein 4 (BRD4), and two selected compounds were successfully synthesized according to the synthetic routes proposed by SynPROTAC. In the following biological experiments, both of them exhibited nanomolar-level degradation activity against BRD4 and potent antiproliferation activity against MV411 tumor cells. These results demonstrate the capability of SynPROTAC to design novel bioactive PROTAC molecules with feasible synthesis routes.