2025/12/30 by Mikaela Farrugia, Paul Helquist, Norrby Per-Ola +1 · 1 voice
Materials Science · Chemistry · Physics and Astronomy · #Machine Learning in Materials Science #Asymmetric Hydrogenation and Catalysis #Advanced Chemical Physics Studies
paper · doi:10.1021/acs.jctc.5c01751
The FUERZA method and its modifications have greatly improved the generation of force field parameters from electronic structure calculations. The QFUERZA method presented here addresses some of the shortcomings of the original method and places it in a comprehensive workflow for the further optimization of the derived force constants within the Q2MM workflow. QFUERZA demonstrated improved accuracy when applied to the generation and full optimization of a ground-state force field for cis -platinum and a transition-state force field for the stereoselecting step of the rhodium-catalyzed hydrogenation of enamides, which combines literature MM3* values with new atom types for the reactive center. QFUERZA outperforms three similarly simple approaches to the derivation of force constants from the literature. Application of the gradient optimizer in the Q2MM workflow on the four different preliminary force fields also demonstrates faster convergence to parameters with excellent agreement with reference data when QFUERZA is used as the starting point. Comparison of QFUERZA-derived parameters for transition-state force fields can identify parameters closely associated with the chemical transition, automating parameter selection for the final stage of transition-state parameter refinement.