Research

Introduction

Computer simulation of the dynamics of biomolecules by the molecular dynamics (MD) technique yields the possibility of describing and understanding the structure-dynamics-function relationships of molecular processes in terms of interactions at the atomic level. Once the reliability of the molecular models, force fields and computational procedures has been established by comparison of simulated properties with known experimental ones, computer simulation can be a very powerful tool to predict molecular properties that are inaccessible to experimental probes. It provides a microscopic picture which may serve to explain observed behaviour of a molecular system.

Simulation of ligand, inhibitor or coenzyme binding opens the way to calculation and prediction of relative binding constants, which is in turn useful in drug design.

Another application is the simulation of DNA-repressor complexes, which gives insight at the atomic level in the possible mechanisms of protein-DNA recognition.

The prediction of energetic and structural changes caused by modification of amino acids in enzymes is of practical value for the engineering of proteins.

The prediction of the spatial structure of proteins using sequence homology with related proteins of known spatial structure would be very useful. The application of MD simulation to derive spatial structure based on atom-atom distance information obtained by NOE-NMR experiments, or based on structure factor amplitude information obtained by X-ray diffraction experiments has become a standard technique in the area of protein structure refinement.

Finally, Molecular Dynamics makes it possible to simulate biological membranes and the behaviour of proteins that are bound to or embedded in membranes.

Search of configuration space

One of the two basic problems in the field of molecular modelling and simulation is how to efficiently search the vast phase or configuration space which is spanned by all possible molecular configurations, for the global low (free) energy regions, which will be populated by the molecular system in thermal equilibrium.

Various techniques to enhance the searching power of MD simulation are investigated: use of soft-core potentials, local elevation of the potential energy surface, etc.

Long-time scale events

A typical MD simulation of a biomolecule in aqueous solution covers a time period of 1 - 10² nsec. This is much too short for a proper description of properties which show a much longer relaxation time. Due to large energy barriers in the potential energy surface, it may take a molecular system a very long time to cross these barriers and so to sample configuration space. Therefore, possibilities to lengthen the time scale of MD simulations are continuously investigated.

Testing and improvement

The reliability of the molecular models, force fields and computational procedures can be established by a comparison of simulated properties with known experimental ones, for various types of molecules: water, chloroform, DMSO, peptides, proteins, DNA, carbohydrates. The introduction of polarisability in the force field for biomolecules is investigated. Methods for the proper treatment of long-range electrostatic interactions are being developed.

Simplification of force fields

When simulating a biomolecule in aqueous solution, the bulk of the computing effort is spent on simulating water. Omitting the solvent degrees of freedom saves a factor of 10-50 in computer time. However, the mean force exerted by the solvent must be retained for a proper treatment of the solute. Various representations of the mean solvation force for biomolecules are investigated and tested. Another way to simplify a force field is to consider only a few degrees of freedom per group of atoms, e.g. an amino acid residue. Such a simplified coarse-grained force field is under development for water, co-solvents and lipids.

Free energy differences

Most chemical quantities of interest, like binding constants, solubilities, adsorption coefficients and chemical potentials, are directly related to the free energy. Since the development of computer simulation methods, it has been attempted to compute the free energy by a variety of statistical-mechanical formulae and procedures. Improvements and alternatives to the existing methodology are being investigated and the developed techniques are applied to complex biomolecular systems.

Electrostatics I

The approximate treatment of electrostatic interactions in computer simulations of (bio-)molecular systems currently represents one of the bottlenecks in the accuracy of these methods. This is because, due to computational costs, simulated systems are restricted to very limited sizes (=1000 nm³), which are typically small compared to the range of electrostatic interactions in these systems. Thus, electrostatic interactions cannot be computed in an exact manner, and uncontrolled approximations can give rise to important artifacts (finite-size effects), which may impair the reliability of many current simulations. A strategy followed in our group to analyze and improve electrostatic schemes for explicit-solvent molecular simulations is to use continuum electrostatics with the goal of understanding, correcting and ultimately eliminating finite-size effects.

Electrostatics II

Because of their magnitude and long-range nature, electrostatic interactions play an important role in determining the properties of (bio-)molecular systems. A combination of methods such as continuum electrostatics, explicit-solvent MD, and a method developed in our group to perform MD at constant pH, are applied to investigate the role of electrostatic interactions in the context of e.g. complexation equilibria, pH-dependent conformational equilibria, protein folding, effect of mutations on protein stability, etc.

NMR structure

The goal of structure determination based on experimental NMR data is to find molecular structures that satisfy the experimental data, such as measured resonance intensities, and have a low energy in terms of a molecular potential energy function. The technique of restraining only the average molecular properties with respect to the measured values is investigated and further developed.

X-ray structure

The application of MD techniques in crystallographic refinement is further developed and tested. As in the NMR case, the application of time-averaged structure factor restraints leads to a better representation of the measurement. The full anharmonic and anisotropic motions of the atoms are taken into account. The problem of obtaining reasonably good initial phases to be used in structure refinement of proteins is tackled using a general low resolution representation of such molecules.

QM simulation

Classical simulation has reached the stage where structural and some thermodynamical properties of biochemical systems can now be reproduced and predicted with a reasonable degree of accuracy. However, where processes involve the breaking and making of chemical bonds or other forms of redistribution of the electron density, classical simulation is inadequate. Here some sort of quantum mechanical treatment of certain degrees of freedom in the molecular system need to be considered. Different approaches of quantum simulation are investigated and tested on simple reactions.

Software development

The growth of the field of computer simulation of fluid-like systems has been made possible by the steady and rapid increase of computing power over the last decades: an order of magnitude every 5-7 years. This trend will continue in the near future, since the present growth of computing power is based on the introduction of parallelism. It is investigated how the possibilities of parallel computation can be most efficiently exploited in MD software for biomolecular systems. Further, the use of specialised hardware to speed up the simulations is investigated.

Applications

A variety of (bio)molecular systems of practical interest is simulated to obtain a microscopic picture of their structure and dynamics, e.g. inhibitor binding to plasmepsin II, structure of lipid bilayers or micelles, the stability of polypeptides in different solvents, etc.