Drug discovery teams can now rank millions of compounds in a few hours using standard CPU instances within their own Amazon Web Services environment with AQPotency, a new quantitative model from SandboxAQ. Unlike traditional methods, AQPotency predicts how strongly candidate compounds bind to protein targets using only a target UniProt ID and ligand SMILES strings, eliminating the need for 3D crystal structures and opening analysis to previously inaccessible targets like membrane proteins.
Deploying AQPotency has been available through Claude via Model Context Protocol since August, and is now also available on Amazon SageMaker, giving teams more flexibility in accessing the model. SandboxAQ notes that it does not collect any metadata about users’ targets or molecules to protect privacy.
AQPotency Predicts Small Molecule Potency via UniProt and SMILES
Drug discovery teams now have a new tool for rapidly assessing potential compounds. AQPotency is available to AWS commercial customers in North America through the AWS Marketplace. This capability extends analysis to targets inaccessible to traditional structure-based methods, including membrane proteins, mutant panels, and proteins that have not yet been crystallized. Computational chemists can evaluate compound security liability, off-target panels and selectivity early in the discovery process, streamlining the screening of large small molecule libraries.
Predictions generated by the model allow teams to prioritize compounds before investing in synthesis, free energy perturbation methods, or wet lab experiments. Deployment is simplified for teams already operating within the AWS ecosystem; customers can subscribe to AQPotency on the AWS Marketplace and deploy it as a SageMaker model package, ensuring confidential data handling throughout the process. Users can interact with the model dynamically via a real-time endpoint or submit large asynchronous jobs for screening millions of ligands.
See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
