AquaGen, a generative AI framework for small-molecule motion.

The key point is simple: molecules are not static. Atoms move, solvent matters, and ignoring that can create false confidence in AI drug discovery predictions.

AquaGen models molecular motion at all-atom resolution with explicit water and periodic boundary conditions, while running an order of magnitude faster than traditional molecular dynamics.

That speed matters because it lets teams test more compounds, refine candidates earlier, and make better decisions before moving into the lab.

What I like most is that this is not positioned as replacing physics. It is a physics-grounded “gray-box” approach that remains interpretable and interoperable with molecular dynamics force fields.

Static structure prediction was one major step. Modeling motion, uncertainty, solvent, and binding behavior is the next one. This is what the next generation of AI drug discovery infrastructure looks like.

Valence Labs / Recursion