SandboxAQ debuts AQPotency, an AI model that screens drugs without a protein structure

SandboxAQ has announced the general availability of AQPotency, an AI system aimed squarely at one of drug discovery’s oldest bottlenecks: deciding which molecules are worth testing when scientists don’t have a detailed 3D map of the protein they’re targeting. Branded as a “Large Quantitative Model” (LQM), AQPotency predicts how strongly a candidate compound will bind to and act on a disease‑related protein.

Traditional in‑silico screening tools usually depend on a solved protein structure. AQPotency, by contrast, is designed to operate even when that structural information is missing or incomplete, using patterns learned from large training datasets to estimate binding strength and functional impact. If its performance holds up in real pipelines, that could open up therapeutic targets that structure‑based methods struggle to reach.

The launch coincides with the broader release of AQCat Adsorption Spin, a companion model for materials and catalyst discovery. Together, they underscore a trend in life sciences and chemistry: R&D teams are increasingly pairing wet‑lab experiments with specialized AI models, not just to speed up individual steps, but to rethink how entire discovery workflows are designed.