Toward a Quantum-Chemical Benchmark Set for Enzymatically Catalyzed Reactions: Important Steps and Insights

We present steps
toward the construction of a quantum-chemical
benchmark set for enzymatically catalyzed reactions comprising differentquite
sizablemodel systems for five reactions. Our work is inspired
by a study of semiempirical molecular-orbital theory methods against
B3LYP-based structures and energies [Kromann et al. PeerJ 2016, 4, e1994]. Taking the same model
systems, we herein demonstrate first how adequate treatments of London
dispersion and basis-set superposition error lead to structural differences
compared to structures obtained at the popular B3LYP/6-31G­(d,p) level.
The changes observed by us have a significant influence on the final
reaction barrier heights (BHs) and energies (REs), which are our main
focus herein. We then proceed and carefully investigate different
strategies to obtain either exact or estimated complete-basis-set
BHs and REs with the help of the DLPNO-CCSD­(T) approach, which offers
an exciting path into exploring larger biochemically relevant structures
at the coupled cluster level. On the basis of our analysis, we present
three strategies for obtaining reliable values. While these recommendations
constitute our main findings and are meant to be used as useful guidelines
for others in the field, we demonstrate their applicability in a preliminary
benchmark set comprising 16 BHs and 12 REs on which we assess 35 density
functional theory (DFT) approximations. This preliminary benchmark
study was used as an indicator for the validity of our approach toward
generating such a benchmark set, based on the fact that we were able
to reproduce major recommendations in the field of modern DFT, such
as the reproduction of the famous Jacob’s Ladder scheme, while
also pointing out subtle differences to previous benchmark studies.
We end our study by critically assessing the still popular choice
to gauge the accuracy of low-level methods against B3LYP data, and
we show the risks of such a strategy, adding to the ever-growing list
of calls to move away from B3LYP as a standard black-box approach.