Benchmarking Density Functional Theory Methods for Metalloenzyme Reactions: The Introduction of the MME55 Set

Deep Mind 21 functional does not extrapolate to transition metal chemistry

The development of density functional approximations stands at a crossroads: while machine-learned functionals show potential to surpass their human-designed counterparts, their extrapolation to unseen chemistry lags behind. Here we assess how well the recent Deep Mind 21 (DM21) machine-learned functional [Science, 2021, 374, 1385-1389], trained on main-group chemistry, extrapolates to transition metal chemistry (TMC). We show that DM21 demonstrates comparable or occasionally superior accuracy to B3LYP for TMC, but consistently struggles with achieving self-consistent field convergence for TMC molecules. We also compare main-group and TMC machine-learning DM21 features to shed light on DM21's challenges in TMC. We finally propose strategies to overcome limitations in the extrapolative capabilities of machine-learned functionals in TMC.

High-Performance Multi-GPU Analytic RI-MP2 Energy Gradients

This article presents a novel algorithm for the calculation of analytic energy gradients from second-order Møller-Plesset perturbation theory within the Resolution-of-the-Identity approximation (RI-MP2), which is designed to achieve high performance on clusters with multiple graphical processing units (GPUs). The algorithm uses GPUs for all major steps of the calculation, including integral generation, formation of all required intermediate tensors, solution of the Z-vector equation and gradient accumulation. The implementation in the EXtreme Scale Electronic Structure System (EXESS) software package includes a tailored, highly efficient, multistream scheduling system to hide CPU-GPU data transfer latencies and allows nodes with 8 A100 GPUs to operate at over 80% of theoretical peak floating-point performance. Comparative performance analysis shows a significant reduction in computational time relative to traditional multicore CPU-based methods, with our approach achieving up to a 95-fold speedup over the single-node performance of established software such as Q-Chem and ORCA. Additionally, we demonstrate that pairing our implementation with the molecular fragmentation framework in EXESS can drastically lower the computational scaling of RI-MP2 gradient calculations from quintic to subquadratic, enabling further substantial savings in runtime while retaining high numerical accuracy in the resulting gradients.

Elucidation of the catalytic mechanism of a single-metal dependent homing endonuclease using QM and QM/MM approaches: the case study of I-PpoI

Homing endonucleases (HEs) are highly specific DNA cleaving enzymes, with I-PpoI having been suggested to use a single metal to accelerate phosphodiester bond cleavage. Although an I-PpoI mechanism has been proposed based on experimental structural data, no consensus has been reached regarding the roles of the metal or key active site amino acids. This study uses QM cluster and QM/MM calculations to provide atomic-level details of the I-PpoI catalytic mechanism. Minimal QM cluster and large-scale QM/MM models demonstrate that the experimentally-proposed pathway involving direct Mg2+ coordination to the substrate coupled with leaving group protonation through a metal-activated water is not feasible due to an inconducive I-PpoI active site alignment. Despite QM cluster models of varying size uncovering a pathway involving leaving group protonation by a metal-activated water, indirect (water-mediated) metal coordination to the substrate is required to afford this pathway, which renders this mechanism energetically infeasible. Instead, QM cluster models reveal that the preferred pathway involves direct Mg2+-O3' coordination to stabilize the charged substrate and assist leaving group departure, while H98 activates the water nucleophile. These calculations also underscore that both catalytic residues that directly interact with the substrate and secondary amino acids that position or stabilize these residues are required for efficient catalysis. QM/MM calculations on the solvated enzyme-DNA complex verify the preferred mechanism, which is fully consistent with experimental kinetic, structural, and mutational data. The fundamental understanding of the I-PpoI mechanism of action, gained from the present work can be used to further explore potential uses of this enzyme in biotechnology and medicine, and direct future computational investigations of other members of the understudied HE family.

Exploring CPS-Extrapolated DLPNO-CCSD(T1) Reference Values for Benchmarking DFT Methods on Enzymatically Catalyzed Reactions

Domain-based local pair natural orbital coupled-cluster singles doubles with perturbative triples [DLPNO-CCSD(T)] is regularly used to calculate reliable benchmark reference values at a computational cost significantly lower than that of canonical CCSD(T). Recent work has shown that even greater accuracy can be obtained at only a small additional cost through extrapolation to the complete PNO space (CPS) limit. Herein, we test two levels of CPS extrapolation, CPS(5,6), which approximates the accuracy of standard TightPNO, and CPS(6,7), which surpasses it, as benchmark values to test density functional approximations (DFAs) on a small set of organic and transition-metal-dependent enzyme active site models. Between the different reference levels of theory, there are changes in the magnitudes of the absolute deviations for all functionals, but these are small and there is minimal impact on the relative rankings of the tested DFAs. The differences are more significant for the metalloenzymes than the organic enzymes, so we repeat the tests on our entire ENZYMES22 set of organic enzyme active site models [Wappett, D. A.; Goerigk, L. J. Phys. Chem. A 2019, 123, 7057-7074] to confirm that using the CPS extrapolations for the reference values has negligible impact on the benchmarking outcomes. This means that we can particularly recommend CPS(5,6) as an alternative to standard TightPNO settings for calculating reference values, increasing the applicability of DLPNO-CCSD(T) in benchmarking reaction energies and barrier heights of larger models of organic enzymes. DLPNO-CCSD(T1)/CPS(6,7) energies for ENZYMES22 are finally presented as updated reference values for the set, reflecting the recent improvements in the method.