conformer-rl 0.1.0

Deep Reinforcement Library for Conformer Generation

conformer-rl

An open-source deep reinforcement learning library for conformer generation.

Documentation

Documentation can be found at https://conformer-rl.readthedocs.io/.

Installation

• We recommend installing in a new conda environment.

  • If you are new to using conda, you can install it here and learn more about environments here
  • Create and activate a new environment:

    $ conda create --name conformerrl # create a new environment
    $ conda activate conformerrl # activate the new environment
    

• Install dependencies

  • Install RDKit

    $ conda install -c conda-forge rdkit
    

  • We recommend installing the dependencies and versions listed in requirements.txt:

    $ pip install -r requirements.txt
    

    The library will most likely still work if you use a different version than what is listed in requirements.txt, but most testing was done using these versions.

• Install conformer-rl

    $ pip install conformer-rl
  

  • If you did not install dependencies using requirements.txt, you will need to manually install Pytorch Geometric here.

• Verify Installation:
  As a quick check to verify the installation has succeeded, navigate to the examples directory and run base_example.py. The script should finish running in a few minutes or less. If no errors ware encountered then most likely the installation has succeeded.

Features

• Agents - conformer_rl contains implementations of agents for several deep reinforcement learning algorithms, including recurrent and non-recurrent versions of A2C and PPO. conformer_rl also includes a base agent interface BaseAgent for constructing new agents.

• Models - Implementations of various graph neural network models are included. Each model is compatible with any molecule.

• Environments - Implementations for several pre-built environments that are compatible with any molecule. Environments are built on top of the modularized ConformerEnv interface, making it easy to create custom environments and max-and-match different environment components.

• Analysis - conformer_rl contains a module for visualizing metrics and molecule conformers in Jupyter/IPython notebooks. The example notebook in the examples directory shows some examples on how the visualizing tools can be used.

Quick Start

The examples directory contain several scripts for training on pre-built agents and environments. Visit Quick Start to get started.