SandboxAQ Launches AI Model to Speed Drug Discovery

SandboxAQ has announced the general availability of AQPotency, an AI-powered model designed to help drug discovery teams predict how strongly potential drug molecules may act on disease targets.

The Large Quantitative Model (LQM) is now available through Claude using the Model Context Protocol (MCP). SandboxAQ said the technology can help researchers screen large numbers of molecules in seconds, allowing them to focus laboratory testing on the candidates most likely to succeed.

In early drug discovery, researchers often need to test many potential molecules before identifying a promising drug candidate. Traditional computer-based methods can be expensive and may require detailed three-dimensional structures of disease targets. Many important biological targets do not have these structures, limiting the use of conventional structure-based approaches.

AQPotency is designed to address this challenge by predicting interactions between molecules and targets without requiring a detailed structural map of the target. The model can rank molecule-target pairs in seconds and can operate on standard computing hardware.

According to SandboxAQ, AQPotency can cost as little as $1 for 1,000 comparisons. Each prediction also includes information about the model’s confidence and whether the target is within the range where the model is considered reliable.

The technology can also be used to investigate potential targets for an existing molecule. Researchers can provide a promising compound and use AQPotency to scan a broad range of proteins, generating a ranked list of potential targets. This could help scientists understand how a molecule produces a particular biological effect and identify new opportunities for drug development.

SandboxAQ said the model has already been used in eight customer programs, with experimental testing supporting its predictions.

Professor Dario R. Alessi of the University of Dundee said the company’s models have helped researchers explore a larger biochemical space while improving the activity and selectivity of potential treatments for Parkinson’s disease.

SandboxAQ has also worked with researchers at Columbia University on the discovery of selective SV2C binders from a large commercial compound library. The work could support the development of new research tools and potential treatments for Parkinson’s disease and other disorders involving dopaminergic signaling.

Andrea Bortolato, Vice President of Drug Discovery at SandboxAQ, said AQPotency provides a faster and scalable way to prioritize compounds without requiring a 3D crystal structure of the target.

The launch adds another AI-based tool to the growing use of machine learning and quantitative models in pharmaceutical research. By helping researchers screen more compounds before laboratory testing, technologies such as AQPotency could reduce the time and cost involved in early-stage drug discovery.

AQPotency is available through Claude via MCP and through the SandboxAQ website. SandboxAQ said availability through Google Cloud Marketplace will follow. The company has also announced general availability of another LQM, AQCat Adsorption Spin, which is designed for catalyst discovery.