Mechanistic Insights into Enzyme Catalysis from Explaining Machine-Learned Quantum Mechanical and Molecular Mechanical Minimum Energy Pathways

Abstract

With the increasing popularity of machine learning (ML) applications, the demand for explainable artificial intelligence techniques to explain ML models developed for computational chemistry has also emerged. In this study, we present the development of the Boltzmann-weighted cumulative integrated gradients (BCIG) approach for effective explanation of mechanistic insights into ML models trained on high-level quantum mechanical and molecular mechanical (QM/MM) minimum energy pathways. Using the acylation reactions of the Toho-1 β-lactamase and two antibiotics (ampicillin and cefalexin) as the model systems, we show that the BCIG approach could quantitatively attribute the energetic contribution in one system and the relative reactivity of individual steps across different systems to specific chemical processes such as the bond making/breaking and proton transfers. The proposed BCIG contribution attribution method quantifies chemistry-interpretable insights in terms of contributions from each elementary chemical process, which is in agreement with the validating QM/MM calculations and our intuitive mechanistic understandings of the model reactions.

Graph-learning guided mechanistic insights into imipenem hydrolysis in GES carbapenemases

“…To obtain accurate energetic profiles, the B3LYP hybrid density functional [51,52] with Grimme's D3 dispersion correction [53] and the 6-31+G * * basis set [54] (B3LYP-D3/6-31+G * * /C36) was used to refine the single point energetics along the DFTB3/3OB-f/C36 optimized MEP geometries. The combination of DFTB3 and B3LYP-D3 levels of theory has been previously proposed and tested on similar systems [30,55,56]. In addition, we compared the MEP configurations optimized at DFTB3/3OB-f/C36 and B3LYP/6-31G * /C36 levels of theory (supporting figure S1).…”

“…Paralleling previous theoretical efforts based on QM/MM are the applications of various machine-learning and deep-learning (DL) techniques in computational enzymology [26][27][28][29][30]. DL has been repeatedly shown as the state-of-the-art computational tool for datasets with high complexity and/or dimensionality, making preferrable for assisting the understanding of enzymatic reactions [31].…”

Constant advance replicas method for locating minimum energy paths and transition states

“…The intriguing work of Brokaw et al 23 represents the first constraint-based CoS method and has demonstrated great efficacy in various applications. [65][66][67][68][69][70][71] We expect the RPCons method as well as the CAR algorithm derived in this paper will facilitate further development of constraint-based CoS methods and their applications to difficult MEP/MFEP problems.…”

Trends in in-silico guided engineering of efficient polyethylene terephthalate (PET) hydrolyzing enzymes to enable bio-recycling and upcycling of PET

“…There have been significant efforts to identify suitable descriptors, and more accurate Machine Learning models in order to incorporate quantum mechanical effects in enzyme engineering. For example, Song and coworkers calculated the minimum energy pathways (MEP) using a multiscale modeling approach [65] . Many initial configurations were obtained using the Molecular Dynamics trajectory.…”