Introduction to molecular simulation

Computer simulations have become an indispensable tool in many areas of science. They can be used to study systems that cannot be solved analytically. They can be employed as “computer experiments” to test theories or to generate new theoretical concepts. They also permit access to levels of detail that are often not accessible to experiments. For example, one has complete control over the initial conditions and can track movements of all particles at all times. A domain where computer simulations have historically enjoyed much success is the study of the properties of materials. Indeed, simulations of crystals and liquids were among the first applications of computer simulation techniques. Molecular dynamics simulations of biomolecular systems – including applications to a wide range of protein properties, such as details of molecular conformations, transmembrane transport, and protein-ligand binding – are rapidly growing in importance [5]. Here, we will provide an introduction to the basics techniques and theory behind computer simulations. More complete descriptions of background theory and key algorithms can be found for instance in [6, 1, 9, 11, 14]. Statistical mechanics is crucial for understanding the theory and techniques behind molecular simulations, and some background knowledge is assumed in this chapter. Many textbooks cover this topic, including [4, 15]. Rapid increases in computer processing power, the emergence of new hardware architectures and simulation algorithms that allow massive parallelization, as well as the development of novel simulation algorithms, together with software packages that facilitate their use have greatly increased the complexity of systems that can be simulated. However, the more difficult the studied system, the more dangerous it becomes to treat simulation software just as a black box. Without deeper understanding of the algorithms and methods employed, it is easy to fool oneself. Our aim is to provide a bit a better appreciation of what is going under the hood when such a simulation package is employed. What are the approximations used? How do these impose limits on validity and applicability? How do I know whether I can extract real physical insight, or have created nothing but a pretty, but potentially misleading, picture or movie? Such questions must never be left aside if one wants to properly distinguish bon fide predictions of real system behavior from simulation artifacts. The most common techniques employed to study thermodynamic properties of systems of particles are Molecular dynamics and Monte Carlo algorithms, which we briefly review. We then discuss some common techniques for speeding-up simulations and overcoming free-energy barriers, and finally provide a list of some of the popular simulation packages for biomolecular systems.