MolecularAI/REINVENT4

AI molecular design tool for de novo design, scaffold hopping, R-group replacement, linker design and molecule optimization.

What it solves

REINVENT 4 is a molecular design tool used for the de novo design of small molecules. It addresses the challenge of creating new chemical structures that meet specific, user-defined property profiles, such as scaffold hopping, R-group replacement, linker design, and general molecule optimization.

How it works

The tool employs a Reinforcement Learning (RL) algorithm to generate molecules that comply with a multi-component score defined by the user. To refine the generation process, Transfer Learning (TL) can be used to pre-train models so that the generated molecules are more closely aligned with a specific set of input molecules.

Who it’s for

It is designed for researchers and chemists working in drug discovery and molecular design who need to optimize small molecules based on specific chemical properties.

Highlights

  • Versatile Design Tasks: Supports de novo design, scaffold hopping, R-group replacement, and linker design.
  • Customizable Scoring: Uses a multi-component scoring system to guide the RL algorithm.
  • Extensible Architecture: Features a plugin mechanism for users to add their own scoring components without modifying the core code.
  • Flexible Hardware Support: Runs on both CPU and GPU, with support for NVIDIA, AMD, Intel ARC, and Apple GPUs via PyTorch.

Related

  • Project
    microsoft/mattergen

    MatterGen is a generative diffusion model for inorganic materials design that can be fine-tuned to generate crystal structures based on specific property constraints.

  • Project
    OTeam-AI4S/ODesign

    An all-atom generative world model for biomolecular interaction design that generates proteins, ligands, and nucleic acids tailored to bind to specific targets.

  • Project
    ReinFlow/ReinFlow

    A policy gradient framework for fine-tuning flow matching policies using online reinforcement learning, enabling efficient optimization of robotic control policies with minimal denoising steps.

  • Project
    deepmodeling/deepmd-kit

    A deep learning package that turns quantum-mechanical reference data into fast, scalable interatomic potentials for molecular and materials science simulations.

  • Project
    NVIDIA/TensorRT-LLM

    An optimization library for LLMs and visual generation models that uses specialized kernels and an efficient runtime to maximize inference throughput and minimize latency on NVIDIA GPUs.