QuantaMind, an AI-Powered Molecular Simulation Platform, Publishes Research in Science Advances

Chinese company Fenzi Zhixin has published research on its self-developed reactive atomistic modeling platform QuantaMind in the journal Science Advances. The work represents a major advance in reactive molecular dynamics simulation, extending near-quantum-chemical accuracy to biological systems of tens of thousands of atoms and timescales of tens of nanoseconds. Further extended tests have achieved simulations of 100,000-atom systems at the hundred-nanosecond scale on a single GPU.

QuantaMind is a reactive machine learning force field framework designed to address the long-standing trade-off among accuracy, speed, and system size in molecular dynamics. Under an NVIDIA A100 80GB environment, the platform was reported to reach advanced inference efficiency, with a computational time of 0.25 seconds per time step for a 100,000-atom reaction system. In validation, the platform completed simulations of a 24,001-atom water system and a 17,792-atom enzyme catalytic system, achieving stable simulations up to 20 nanoseconds. In a PET hydrolase test, QuantaMind simulated a full catalytic cycle, with atomic force component correlation coefficients exceeding 0.99 against DFT references.

The company says QuantaMind can support drug development and enzyme engineering by enabling a “compute first, design then” workflow. Its capabilities are planned to be integrated into the AI-native bio-economy operating system MoleculeOS.