Skip to content

What is OpenAD?

OpenAD is a technical framework for molecular discovery. OpenAD is a new approach for providing easy access to machine learning tools for computational chemists. Since it is so unique, explaining what it is can sometimes be challenging. Let’s break down what “framework for molecular discovery” means.

“Molecular discovery” refers to the tools that are used within OpenAD. For the most part these tools consist of machine learning models that are built for the purpose of predicting chemical reactions and properties. Using these models, you can do things like retrosynthesis prediction (i.e., taking a molecule and working backwards to predict the reagents and reactions that created it) and forward prediction, where you give a model a set of reagents and ask it to predict the final products. You can also use OpenAD to compute specific properties of molecules (e.g., molecular weight, number of aromatic rings, number of hydrogen acceptors or donors), proteins (e.g., charge density, protein weight), crystals (e.g., absolute energy, Poisson ratio, formation energy), and properties generated from MolFormer models. OpenAD gives you a convenient and consistent environment to deploy and use these machine learning tools.

Normally, this type of molecular discovery requires work from different technical disciplines, like cloud computing and system administration. You have to set up a lot of different tools and platforms just to get started. Maintaining and updating these programs also becomes a challenge.

To use an analogy, let’s say you’re a new executive chef of a new and expensive Italian restaurant. However, you are surprised to learn that your new job includes responsibility for constructing the physical building, kitchen, and dining area. You would much rather spend your time working with your preferred tools and ingredients. Much like a chef with a restaurant space already built, you can use OpenAD as the framework for organizing and using your tools to create and analyze chemical reactions however required. You can be creative and explore results with many different models, or you can define model workflows that can be used on an industrial scale. All of this is possible with OpenAD.

OpenAD is a game-changing, open-source framework that democratizes IBM’s research in machine learning and advanced compute for molecular discovery. This powerful tool provides a simple, command-driven solution that accelerates your molecular research by delivering:

  • Easy-to-use AI models: Deploy scientific AI models directly into your Jupyter or application workflows using a domain-specific command language.
  • Molecular data awareness: Manage, subset, and visualize molecule sets in notebooks or applications with seamless integration of services like PubChem, Deep Search, and RXN.
  • Federated model and information services: Integrate a wide range of Model and Information services into your Discovery workflow, enriching your research with patent, research paper, and reaction prediction information.
  • Option of Secure Service Delivery: While OpenAD is a open source project it has the ability for you to host and securely lock away your model assets that contain proprietary or confidential information behind a secure firewall, whether that is deployed and hosted by IBM or within your own environment, making it easy to transition from open source to confidential deployments.
  • Open Source Platform: Strategic alignment with the AI Alliance to provide transparent, safe, reliable, and open source tools for science.

With OpenAD, you can combine resources from IBM, third-party projects, open-source libraries, or custom tool kits into a single pipeline. This flexibility enables you to:

  • Orchestrate workflows: Catalog user-defined or open-source model services, launch and manage model services, and lower costs of managing services.
  • Build new pipelines: Design OpenAD as the foundation for building new workflows from scratch.

You can learn more about OpenAD and download the free OpenAD Toolkit by visiting https://accelerate.science/projects/openad. 

Read On