Abstract:A quantum mechanics/molecular mechanics (QM/MM) approach is a broadly used tool in computational enzymology. Treating the QM region with a high-level DFT method is one of the important branches. Here, taking leaf-branch compost cutinase-catalyzed polyethylene terephthalate depolymerization as an example, the convergence behavior of energy barriers as well as key structural and charge features with respect to the size of the QM region (up to 1000 atoms) is systematically investigated. BP86/6-31G(d)//CHARMM and M06-2X/6-311G(d,p)//CHARMM level of theories were applied for geometry optimizations and single-point energy calculations, respectively. Six independent enzyme conformations for all the four catalytic steps (steps (i)-(iv)) were considered. Most of the twenty-four cases show that at least 500 QM atoms are needed while only two rare cases show that similar to 100 QM atoms are sufficient for convergence when only a single conformation was considered. This explains why most previous studies showed that 500 or more QM atoms are required while a few others showed that similar to 100 QM atoms are sufficient for DFT/MM calculations. More importantly, average energy barriers and key structural/charge features from six conformations show an accelerated convergence than that in a single conformation. For instance, to reach energy barrier convergence (within 2.0 kcal mol-1) for step (ii), only similar to 100 QM atoms are required if six conformations are considered while 500 or more QM atoms are needed with a single conformation. The convergence is accelerated to be more rapid if hundreds and thousands of conformations were considered, which aligns with previous findings that only several dozens of QM atoms are required for convergence with semi-empirical QM/MM MD simulations. Both QM region sizes and conformation numbers have an impact on modelling enzyme reactions. The convergence behavior of energy barriers and features with respect to the QM region size can be accelerated by considering multiple conformations.

Understanding the Effectiveness of Enzyme Pre-Reaction State by a Quantum-Based Machine Learning Model

Quantitative prediction of enantioselectivity of Candida antarctica lipase B by combining docking simulations and quantitative structure-activity relationship (QSAR) analysis

Unraveling the energetic significance of chemical events in enzyme catalysis via machine-learning based regression approach

Predicting the Stereoselectivity of Chemical Transformations by Machine Learning

In silico prediction of metabolic stability for ester-containing molecules: Machine learning and quantum mechanical methods

Predicting the stereoselectivity of chemical reactions by composite machine learning method

Accelerating the optimization of enzyme-catalyzed synthesis conditions via machine learning and reactivity descriptors.

Machine-Learning-Assisted Free Energy Simulation of Solution-Phase and Enzyme Reactions

Machine Learning Prediction of Structure‐Performance Relationship in Organic Synthesis

Enhanced stereodivergent evolution of carboxylesterase for efficient kinetic resolution of near-symmetric esters through machine learning

Modern Semiempirical Electronic Structure Methods and Machine Learning Potentials for Drug Discovery: Conformers, Tautomers, and Protonation States

Exploration and Evaluation of Machine Learning-Based Models for Predicting Enzymatic Reactions

Quantum chemical modeling of enzymatic reactions: The case of decarboxylation

Machine Learning Identifies Chemical Characteristics That Promote Enzyme Catalysis

Prediction of transition state structures of gas-phase chemical reactions via machine learning

EnzyKR: a chirality-aware deep learning model for predicting the outcomes of the hydrolase-catalyzed kinetic resolution

Enantioselectivity prediction of pallada-electrocatalysed C–H activation using transition state knowledge in machine learning

Highly accurate prediction of reaction performance and enantioselectivity with a uniform machine learning protocol

Impacts of QM region sizes and conformation numbers on modelling enzyme reactions: a case study of polyethylene terephthalate hydrolase

Universal Descriptors of 'Quasi Transition States' for Small-Data-Driven Asymmetric Catalysis Prediction in Machine Learning Model

A machine learning approach for reliable prediction of amino acid interactions and its application in the directed evolution of enantioselective enzymes