SandboxAQ

Be the first. The limits of today’s methods should never become the limits of your scientific ambition. SandboxAQ’s Large Quantitative Models are a new class of AI combining machine learning with physics-based simulation to extend what you can rigorously investigate.

We recently shared news of our Nature Communications paper with Baylor College of Medicine. C&EN has also covered the work. Molecular glues pull two proteins together and prompt the cell to destroy a target it would otherwise leave alone. That mechanism opens up targets the field has long considered undruggable, including VAV1. The C&EN piece focuses on how we found these molecules. Baylor screened a small library of chemical fragments using high throughput proteomics and identified two early hits. SandboxAQ then applied AQFEP, our physics based free energy platform, to model how those compounds engage VAV1 and to show chemists where to improve them. Both teams told C&EN that starting with fragments makes molecular glue discovery more affordable and more accessible than conventional approaches. Congratulations to Rae (Rui) Qi, Ly Le, and Andrea Bortolato, and to our collaborators at Baylor. Read the C&EN article: https://lnkd.in/eifJNG7k Read the paper: https://lnkd.in/ekxSiKV7 Baylor's press release: https://lnkd.in/efsTe_dw

Congratulations to our SandboxAQ colleagues Rae (Rui) Qi, Ly Le, and Andrea Bortolato on an exciting milestone in our ongoing collaboration with Baylor College of Medicine — a new paper published in Nature Communications! Read the paper here: https://lnkd.in/ekxSiKV7 VAV1 is a key target in hematological malignancies and autoimmune diseases and has long been considered challenging to drug. This collaboration is a powerful example of how experimental biology, AI, and physics-based approaches can come together to tackle that challenge. Baylor’s high-throughput proteomics and experimental validation, combined with AI-based ternary-complex structure modeling and FEP, helped identify and characterize molecular glue degraders for VAV1. This integrated approach identified a non-canonical RT-loop degron and enabled prospective ranking of molecular glue analogs based on predicted ternary-complex cooperativity. Our SandboxAQ team contributed AQFEP, our physics-based free-energy platform, to characterize and rank molecular glue candidates, with predictions consistent with experimental degradation results. We’re proud to see computational science and experimental biology come together in this collaboration, and we look forward to continuing our work with Baylor to advance molecular glue discovery. See Baylor’s Press Release at: https://lnkd.in/efsTe_dw

Millions of compounds ranked against your target in seconds. About $1 per thousand. No crystal structure, no GPU. Watch it work, then sign up to try it yourself.

Virtual screening can be computationally intensive and difficult to scale. See how Claude and AQPotency make it easier to rank compounds by predicted potency and selectivity, without writing code. Watch the demo, then sign up to explore AQPotency yourself.

AQPotency is a Large Quantitative Model for potency prediction and virtual screening that ranks protein-ligand pairs by predicted binding strength (pIC50) without requiring a 3D crystal structure of the target.