Assessment of Strong-Correlation Corrected Range-Separated Local Hybrid Functionals for Metalloenzyme Reactions.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Theoretical Chemistry

Background:

• Accurately modeling metalloenzyme active sites is crucial but challenging for density functional theory (DFT).
• Existing benchmarks like MME55 provide high-level reference data for large metalloenzyme models.
• Previous top-performing functionals include MPW1B95-D3(BJ) and ωB97M-V.

Purpose of the Study:

• To evaluate range-separated local hybrids (RSLHs) and strong-correlation corrected RSLHs (scRSLHs) for metalloenzyme active site modeling.
• To assess the performance of scRSLHs on enzyme reaction benchmarks (ENZYMES22 and ECR20).
• To provide critical assessments of scRSLHs' practical utility, including computational cost and implementation details.

Main Methods:

• Tested RSLHs and scRSLHs against the MME55 benchmark with DLPNO-CCSD(T)/CBS reference values.
• Evaluated functionals on ENZYMES22 and ECR20 datasets, deriving new CCSD(T)/CBS reference values for ECR20.
• Analyzed grid dependence, computational cost, and parallel scaling of scRSLHs in Turbomole.

Main Results:

• The scRSLH ωLH23tdB-D4 achieved the lowest mean absolute error (MAE) of 2.46 kcal/mol on MME55.
• ωLH23tdB-D4 also showed excellent performance on ENZYMES22 (MAE 1.45 kcal/mol) and ECR20 (MAE 0.63 kcal/mol).
• Accuracy improvements stem from functional design beyond the real-space nondynamical correlation model.

Conclusions:

• scRSLHs represent a significant advancement in DFT for metalloenzyme active site modeling.
• ωLH23tdB-D4 is a top-performing functional for these challenging systems and reaction energy benchmarks.
• This study provides valuable insights into the practical application and performance of emerging scRSLH functionals.