Shanghai – September 17, 2026 -- MoleculeMind, an AI-native bioengineering infrastructure company, has had research on its reactive atomistic modeling platform QuantaMind published in Science Advances, demonstrating Density Functional Theory (DFT)-level accuracy in simulating 10,000-atom biomolecular systems over tens-of-nanosecond timescales.
QuantaMind becomes first platform to simulate a complete enzyme reaction cycle
The study shows QuantaMind enabling AI to observe the full atomic-scale process of proton transfer, bond breaking and formation, and complete enzyme catalytic cycles. In industrial deployments, the platform has been scaled further to model reactive systems comprising hundreds of thousands of atoms.
Benchmark tests show 100,000-atom simulation completed in 0.25 seconds
QuantaMind, built as a transition-state-centered machine learning force field (MLFF), incorporates non-equilibrium conformations and DFT method classification embeddings to unify quantum chemistry data from different functionals and basis sets into a single training framework. Latest benchmark results show it can run a single time-step simulation of a 100,000-atom reactive system in 0.25 seconds, a result MoleculeMind describes as state-of-the-art among leading MLFFs.
Antibody project records 62-fold faster dissociation rate at lower pH
In a drug discovery application involving pH-sensitive antibodies engineered for extended half-life, QuantaMind was applied alongside other AI design models. Experimental testing on one resulting candidate showed a dissociation rate at pH 6.0 approximately 62 times faster than at pH 7.4, illustrating how the platform links molecular design to measurable reaction behavior.
MoleculeMind positions QuantaMind as infrastructure linking design, simulation and validation
The company has also applied QuantaMind to enzyme engineering, using it to uncover reaction mechanisms of complex catalytic enzymes and guide mutation selection, shifting design work from trial-and-error toward computationally guided validation. Xu Jinbo, Founder of MoleculeMind, said AI-powered molecular R&D has largely focused on molecular structure and design, while QuantaMind was built to make reaction processes traceable, mechanisms analyzable and conclusions verifiable.