Molecular Dynamics Simulation in Biomacromolecule Characterization

The application of molecular dynamics simulation in the characterization of biological macromolecules provides a powerful tool for the study of complex molecular behavior in life processes. Biological macromolecules, such as proteins, nucleic acids, lipids, and sugars, are highly structurally complex and dynamic, and MD simulations are able to reveal the structural changes of these molecules under different conditions, their functional mechanisms, and the details of their interactions with other molecules by tracking their movements at the atomic level. MD simulations not only help to understand the properties of molecules on static structures but also provide dynamic information about how molecules evolve over time, which is critical for basic life science research and drug development.

Overview of Biomacromolecule Characterization

Biological macromolecules (such as proteins, nucleic acids, and polysaccharides) play a key role in biological processes. Therefore, it is urgent to accurately characterize the structure, function, and interactions of these molecules to help understand their biological mechanisms and applications. The characterization of biological macromolecules is a process of analyzing their molecular structure, dynamic properties, interaction modes, and functional properties by using various experimental techniques and computational tools. With the development of computational technology, computational chemistry, and molecular dynamics simulation have become important tools for the characterization of biological macromolecules. These simulations can not only predict the structural and dynamical properties of large molecules but also simulate their interactions with other molecules, such as small molecule drugs, proteins or DNA, thus providing a theoretical basis for drug design and disease research.

Biological macromolecules play a key role in life activities, and their structural, functional and dynamic properties are interrelated to determine their role in living organisms. Accurate characterization of biomacromolecules is essential for a deeper understanding of life mechanisms, revealing the causes of disease, and advancing drug development. Through the combination of various experimental and computational methods, we can obtain information about the three-dimensional structure, dynamic behavior, and interactions of these molecules to optimize drug design and improve therapeutic effectiveness. The comprehensive characterization of biomacromolecules not only provides theoretical support for basic scientific research but also opens up new directions for clinical medicine and drug development.

Application of MD in Protein

It is difficult to fully discern the detailed features of atomic motion with experimental methods, and shortly after determining the first protein structure, computers were used to perform MD simulations, first of small globular proteins in a vacuum, followed by larger molecules and molecular complexes under more realistic solvent conditions. The first very short protein simulation does show associated motion on picosecond timescales. An early example of MD simulations linking protein dynamics to function is myoglobin, where oxygen cannot reach binding sites on heme groups if the protein structure is static. Soon, researchers began implementing many methodological developments, such as enhanced sampling, QM/MM, and free energy calculations, which are incorporated into widely used MD simulation programs today.

In the study of proteins, MD simulations help researchers delve into the folding process, stability, functional sites, and interactions with ligands of proteins. By simulating the dynamic behavior of proteins in different environments, MD can reveal how proteins fold from disorder to order, and their transition states during this process. In addition, MD simulation can accurately analyze the binding affinity and binding pattern of protein-ligand, thus providing a theoretical basis for drug design, molecular recognition, and targeted therapy.

Protein Structure Prediction

Homology Modeling and Refinement

The gap between the discovery of new protein sequences and the acquisition of detailed structural information through X-ray diffraction or NMR will remain. For this reason, there is an urgent need for theoretical methods to predict protein structures from sequences. Current attempts to improve homology models to correct for inherent errors when using template methods are usually based on energy minimization, limited conformational sampling using molecular dynamics combined with detailed force fields, or more extensive sampling using simplified force fields.

In the work of Fan H et al., the use of molecular dynamics simulations using atom-based empirical force fields in explicit solvents and for refining de novo or homologically generated protein structures has been investigated. They found that MD simulations using explicit representations of the protein and solvent environment on timescales of 10-100 nanoseconds were useful for refining protein models.

Dynamics of Protein Folding

Understanding how and how quickly proteins move from an unfolded state to a naturally folded structure remains a huge challenge for computer simulations. Force field models in molecular dynamics simulations or Monte Carlo (MC) conformation sampling methods typically focus on capturing the stable conformation and final structure of a protein. MD is based on the numerical integration of the classical Newtonian equations of motion for all atoms in the system, whose interactions are described by parameterized empirical potential functions (or force fields) to capture atomic interactions and fluctuations as realistically as possible. Bond interactions include bond stretching, angular bending, and dihedral angular torsion, and are described by harmonics or other simple potentials. Non-bonded interactions include the van der Waals contribution (described by the Lennard-Jones potential) and the electrostatic force between charged atoms calculated using Coulomb's law.

Figure 1: Protein folding process. (Rizzuti B et al,2013)

Rizzuti B et al. proposed FoldPAthreader, a protein folding pathway prediction method that uses a novel folding force field model by exploring the intrinsic relationship between protein evolution and folding in the known protein world. In addition, the folding force field is used to guide Monte Carlo conformation sampling by exploring potential intermediates that drive protein chain folding into its natural state. On 30 sample targets, FoldPAthreader successfully predicted 70% of proteins whose folding pathways were consistent with biological experimental data.

Study on Enzymatic Catalytic Mechanism

Hybrid Quantum Mechanics/Molecular Mechanics (QM/MM) Simulations

Mahurat et al. investigated the mechanism of type II thymidine kinase Thermotoga maritima TmTK. By using controlled MD (SMD) and umbrella sampling QM/MM MD (US), we described the reaction mechanism and characterized the transition state structure. The free energy barriers of all mechanistic steps of phosphorylation reactions were estimated in an attempt to find the barriers to predicting enzyme-catalyzed reactions.

Allosteric Effects

Allosteric drugs provide a new approach to modern drug design. However, identifying mysterious allosteric sites is a difficult challenge. Following the allosteric properties of residual-driven conformational transitions, Chen, X. et al. propose a state-of-the-art computational pipeline by developing a residual-driven intuitive mixing machine learning (RHML) model combined with molecular dynamics (MD) simulation, through which allosteric sites and allosteric regulators can be effectively identified and their regulatory mechanisms revealed.

Protein-Ligand Interaction

MD simulations provide a detailed dynamic view of the interactions between proteins and small molecule ligands. MD simulations can observe how ligands bind to the active site of the protein and analyze the stability, affinity, kinetic properties, and binding patterns of this binding. This information is critical for drug discovery and optimization, especially in drug design, where MD simulations help to understand the mechanism of interaction between ligands and targets, thereby optimizing the structure of ligands to improve drug specificity and biological activity. For details, see How to Analyze Results from Molecular Dynamics Simulations.

Application of MD Simulation in Nucleic Acid

Molecular mechanics (MM), on which molecular dynamics is based, cannot capture the polarization of DNA components caused by intramolecular and intermolecular interactions, so research in this area has not been fully developed.

In addition to the application of MD simulation to proteins, it can also reveal the fluidity, phase behavior and interaction mechanism of lipid bilayers and membrane proteins. MD simulations can help us understand how membrane proteins are embedded in membranes and how they transport molecules or transmit signals within them, providing new ideas for the design of drug targets.MD simulation is a valuable source of nucleic acid structure information because it includes solvation at the atomic level as well as thermal effects and allows for a wide range of conformational sampling. In fact, over the past decade, there have been efforts to improve or develop increasingly reliable nucleic acid force fields. Unfortunately, MD simulations of medium-sized single strands are scattered all over the place, and some studies of ss-DNA and RNA sequences are primarily aimed at testing the efficiency of the algorithms and the quality of the force fields used in the simulations, rather than providing structural information.

Application of MD Simulation to Other Biological Macromolecules

Studies on Lipids and Membrane Proteins

A classic example is the MD study by Melanie P. Muller et al., in which proteins were simulated in the presence of dominant lipid and lipid-protein interactions and their structural, kinetic, or functional effects were analyzed and reported. An overview will be given of the main simulation techniques used in biofilm computing research, namely atoms (all-atom (AA) or associative atoms (UA)), coarse-grained (CG), and multi-scale characterization and modeling techniques for embedding/inserting membrane-associated proteins into the lipid bilayer. The results obtained by simulation are then detailed in two subsequent sections, divided into the interactions between lipids and integrated membrane proteins and peripheral membrane proteins. Then, we will discuss specific lipids that play a special role in regulating protein structure and function. Finally, the effects of proteins captured by simulation on membrane structure will be reviewed.

Integrating MD with Machine Learning

Enhanced sampling techniques in molecular dynamics simulations are essential for exploring the vast conformational space of biomolecules and capturing their dynamic behavior on biologically relevant timescales. The integration of artificial intelligence, machine learning, and quantum computing can significantly improve these technologies, resulting in more efficient algorithms and more comprehensive exploration strategies.

AI and ML can change the enhanced sampling by introducing adaptive algorithms that adjust the sampling parameters in real-time based on the results of previous simulations. This adaptive sampling allows the system to "learn" from the accumulated data, focusing computational resources on exploring less understood or more critical regions of the conformation space.

The integration of artificial intelligence, machine learning, and quantum computing with molecular dynamics simulations is catalyzing a revolution in computational biology, improving the accuracy and efficiency of simulations. The integration of artificial intelligence and quantum computing with MD simulations offers insightful and stimulating improvements to the understanding of molecular mechanisms, but it may introduce new issues related to data quality, model interpretability, and computational complexity.

Figure 2: Integration of advanced technologies, methods and applications. (Lappala, Anna et al, 2024)