AQPotency is accessible through Claude via the Model Context Protocol (MCP), as well as through the SandboxAQ website, with plans for listing on Google Cloud’s Marketplace.
It enables drug discovery teams to computationally assess and rank how strongly candidate molecules are likely to interact with specific disease targets.
Unlike established computer methods that require detailed structural data about the target molecule, AQPotency does not rely on existing structural maps.
This capability allows researchers to evaluate potential drug compounds even for targets that lack structural information, which has typically limited access to traditional computational screening.
SandboxAQ stated that early-stage drug discovery often involves selecting which molecules to assess experimentally, a process that can be both costly and time-consuming if unsuitable candidates are chosen.
Existing computational narrowing techniques are described as slow and expensive, and restricted to cases where a detailed structure is available.
AQPotency addresses these shortcomings by providing rapid ranking of molecule-target pairs and delivering results in seconds using standard computing resources.
The model reports both the predicted activity and a confidence assessment, indicating whether a target falls within its reliable performance range.
The tool is also designed to support analysis in the opposite direction.
Researchers can input a molecule and receive a prioritised list of proteins that the compound may interact with, which could be useful when a molecule demonstrates effects whose mechanisms are not yet clear.
SandboxAQ drug discovery vice-president Andrea Bortolato said: “AQPotency has given us and our customers a faster, scalable and reliable way to prioritise compounds in the workflows we already run, without needing a 3D crystal structure of the target. This opens up programmes that structure-based methods simply couldn’t reach.
“The confidence intervals make the output actionable for biopharma companies, and the model has already been successfully used in eight customer programmes with experimentally validated impact.”