Improved sampling techniques for the simulation of biocatalytic reaction mechanisms

Hulm, Andreas Valentin (2025): Improved sampling techniques for the simulation of biocatalytic reaction mechanisms. Dissertation, LMU München: Faculty of Chemistry and Pharmacy

Abstract

In recent years, the dramatic improvement of computer hardware, as well as remarkable algorithmic advances, enabled accurate computer simulations of ever-growing molecular systems. Nowadays, with the help of composite quantum-mechanical/molecular-mechanical (QM/MM) methods, it is possible to model 'electronic' processes, like chemical reaction mechanisms that involve covalent bond rearrangements, in extended biological macromolecules with thousands of atoms. Such systems are frequently characterized by remarkable structural flexibility and potential energy surfaces full of local minima. Therefore, to enable the calculation of accurate ensemble properties, representative sampling of their configurational landscape is required, usually employing molecular dynamics (MD) simulations. However, it is crucial that the trajectories obtained are sufficiently long to resolve rare events, i.e., phenomena that occur on macromolecular timescales and are beyond the reach of conventional all-atom molecular dynamics (MD) simulations. Hence, importance sampling techniques are applied to accelerate transitions between metastable states. In addition to accelerating sampling, such methods must provide accurate estimates of reaction free energy differences and kinetic rates of chemical transitions, such that the most likely reaction mechanisms can be found. In this dissertation, several improvements of importance sampling techniques are presented. The focus lies on highly efficient algorithms, which are suitable for use together with an accurate QM/MM treatment of the electronic structure based on density functional theory (DFT), which still limits MD simulations to timescales of only a few hundred picoseconds due to their computational demand. The result is a versatile toolbox of sampling algorithms, which is made publicly available in the open-source adaptive-sampling Python package, that is highly useful for modeling intricate biocatalytic reaction mechanisms in explicit protein environments. This is demonstrated by its application to the simulation of challenging enzymatic systems that catalyze intricate biochemical transitions, such as ribonucleic acid (RNA) modification, adenosine triphosphate (ATP) hydrolysis, or long-range biological protonation dynamics.

• Item Type: Theses (Dissertation, LMU Munich)
• Keywords: Chemistry, Biocatalysis, Molecular modeling
• Subjects: 500 Natural sciences and mathematics; 500 Natural sciences and mathematics > 540 Chemistry and allied sciences
• Faculties: Faculty of Chemistry and Pharmacy
• Language: English
• Date of oral examination: 15. October 2025
• 1. Referee: Ochsenfeld, Christian
• Signature of the printed copy: 0001/UMC 31600