Link-INVENT: Generative Linker Design with Reinforcement Learning

Abstract

In this work, we present Link-INVENT as an extension to the existing de novo molecular design platform REINVENT. We provide illustrative examples on how Link-INVENT can be applied on fragment linking, scaffold hopping, and PROTACs design case studies where the desirable molecules should satisfy a combination of different criteria. With the help of Reinforcement Learning, the agent used by Link-INVENT learns to generate favourable linkers connecting molecular subunits that satisfy diverse objectives, facilitating practical application of the model for real-world drug discovery projects. We also introduce a range of linker-specific objectives in the scoring function of REINVENT. The code is freely available at https://github.com/MolecularAI/Reinvent.

Docking-based generative approaches in the search for new drug candidates

“…Link-INVENT 47 was designed for the fragment-based drug discovery problem: on the basis of a batch of fragments, the encoder-decoder proposes whole linked molecules to be optimized with RL towards best docking (Glide used). In the case of complex multi-parameter optimization (MPO), curriculum learning (CL) might be favorable 48 : divide optimization into several production objectives, e.g.…”

Advancing Targeted Protein Degradation via Multiomics Profiling and Artificial Intelligence

“…Other specialized methods focus on fragment elaboration, which could be employed downstream of proteome-wide fragment screening experiments via aforementioned chemoproteomics. Moreover, existing platforms have been versioned to design optimal linkers connecting the two molecular subunits in a PROTAC . Design of PPI inhibitors has also been attempted, which may set the stage for the design of MGDs.…”

“…Moreover, existing platforms have been versioned to design optimal linkers connecting the two molecular subunits in a PROTAC. 163 Design of PPI inhibitors has also been attempted, 164 which may set the stage for the design of MGDs. As more degraders are discovered, it is likely that dedicated generative models will flourish, borrowing principles apprehended in generalistic models 165 and fine-tuning them to the degrader-specific tasks.…”

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

“…Further, a gated graph neural network (GGNN) outperforms molecular graph generation in deep generative models , and demonstrates the practical structure formation in drug design. , Many approaches use two-dimensional (2D) SMILES-based chemical graphs embedded in low-dimensional space to generate new molecules by perturbing the hidden values of the sampled atoms. − These studies are missing the nature of molecular shape and the three-dimensional (3D) information, which may considerably differ from the starting point of structure design. Another recent popular deep neural network drug design is in the fragment linking technique.…”