REINVENT4 Molecular Discovery App Now Available Via OpenAD

The REINVENT4 molecular discovery tool (https://github.com/MolecularAI/REINVENT4) has recently been incorporated into the Open Accelerated Discovery (OpenAD) framework by use of the OpenAD Service Utilities.  The benefits of this new capability include ease of setup and use, no need to purchase and maintain expensive hardware to run machine learning models, easier scalability on cloud hosting platforms, and easier collaboration with other researchers.

In this blog post, we provide a summary of

1. the OpenAD framework, which includes an OpenAD Toolkit for users and  the OpenAD Service Utilities for developers, and
2. the deployment of the REINVENT4 molecular discovery tool within then OpenAD framework.

What is OpenAD?

OpenAD is a framework for hosting and utilizing machine learning tools for molecular discovery. The hosting of molecular discovery tools within OpenAD allows researchers to easily collaborate with others by sharing their tools and allowing other researchers to access tools via the OpenAD Toolkit (https://github.com/acceleratedscience/open-ad-toolkit) or other client software. Also, individual tools can be incorporated into a  workflow, where the output from one tool can be used as the input for another.

The OpenAD Service Utilities are part of OpenAD and allow developers an easy way to build and deploy a REST API to provide access to almost any molecular discovery tool.   With the OpenAD Service Utilities, developers can easily build and deploy a REST API to access machine learning tools specific to molecular discovery (e.g., identifying molecular properties or generating molecular structures).  The most impressive feature of the OpenAD Service Utilities is that they can be used to incorporate virtually any molecular discovery tool into the OpenAD framework. 

REINVENT4

We have used the OpenAD Service Utilities to deploy a REST API for the REINVENT4 molecular discovery tool set (https://github.com/MolecularAI/REINVENT4).

REINVENT is a molecular design tool for de novo design, scaffold hopping, R-group replacement, linker design, molecule optimization, and other small molecule design tasks. REINVENT uses a Reinforcement Learning (RL) algorithm to generate optimized molecules compliant with a user defined property profile defined as a multi-component score. Transfer Learning (TL) can be used to create or pre-train a model that generates molecules closer to a set of input molecules. A paper describing the software has been published as Open Access in the Journal of Cheminformatics (https://link.springer.com/article/10.1186/s13321-024-00812-5).

REINVENT4 is a command line Python application that includes machine learning models used during its analyses. Using the OpenAD Service Utilities, we have taken this command line app and made it available via a REST API, providing researchers with significant benefits:

• The configuration settings used by the REST API are identical to the ones used by the command line tool, making it easier for researchers to switch from the command line to the REST API.
• Another obvious benefit of the REST API is that researchers do not need to maintain expensive GPU hardware in order to utilize REINVENT4.
• The wrapper code for REINVENT4 can be containerized and deployed on RedHat OpenShift or other host platforms. This makes it easier to scale up the use of REINVENT4 without the need for more expensive hardware.
• The REST API makes it easier to collaborate with researchers across the globe, while ensuring that all collaborators are using the same versions of the same models.

The code for the REINVENT4 OpenAD Service Utilities wrapper and a tutorial on its use will be available very soon. When released we will announce it here on this blog.

For more information on OpenAD and IBM Accelerated Discovery offerings, visit https://accelerate.science.