Fragment Linker Prediction Using the Deep Encoder-Decoder Network for PROTACs Drug Design

Abstract Drug discovery and development pipeline is a prolonged and complex process and remains challenging for both computational methods and medicinal chemists. Deep learning has shed light on various fields and achieved tremendous success in designing novel molecules in the pharmaceutical industry. We utilize state-of-the-art techniques to propose a deep neural network for rapid designing and generating meaningful drug-like Proteolysis-Targeting Chimeras (PROTACs) analogs. Our method, AIMLinker, takes the structural information from the corresponding fragments and generates linkers to incorporate them. In this model, we integrate filters for excluding non-druggable structures guided by protein-protein complexes while retaining molecules with potent chemical properties. The novel PROTACs subsequently pass through molecular docking, taking root-mean-square deviation (RMSD), the change of Gibbs free energy (Δ G binding ), and relative Gibbs free energy (ΔΔ G binding ) as the measurement criteria for testing the robustness and feasibility of the model. The generated novel PROTACs molecules possess similar structural information with superior binding affinity to the binding pockets in comparison to existing CRBN-dBET6-BRD4 ternary complexes. We demonstrate the effectiveness of AIMLinker having the power to design compounds for PROTACs molecules with better chemical properties.