Required cookies

This website uses cookies necessary for its operation in order to provide the user with content and certain functionalities (e.g. language selection). You have no control over the use of these cookies.

Website visitor statistics

We collect visitor statistics on the use of the site. The data is not personally identifiable and is only stored in the Matomo visitor analytics tool managed by CSC.

By accepting visitor statistics, you allow Matomo to use various technologies, such as analytics cookies and web beacons, to collect statistics about your use of the site.

Change your cookie choices and read more about visitor statistics and cookies


Simple models (think of balls connected by springs) can be used to study diffusion, solvation and evaporation, but quite large model systems (millions of atoms or a cube of the size of some tens of nanometers) can be used. Another trick to look at larger systems is to sacrifice atomic detail and lump several atoms to one interacting model particle – this method is called coarse graining.

Looking at bond breaking, like in chemical reactions or electronic transitions, requires modelling electrons explicitly and hence using quantum mechanics. It is computationally expensive and only much smaller systems – and lengths of time – can be studied. Another recent innovation is using machine learning to define how the atoms interact. Miguel Caro et al., from the Aalto university has cracked a long standing mechanistic problem with such a method by being able to look at a model decades bigger and for longer times than before.

Over the years great advances in theoretical models, algorithmic implementations in simulation codes and the speed up of computers have enabled new and ever more useful simulations to be done. Molecular dynamics has been for years a tool in drug design and other molecular property prediction, synthetic chemistry, catalyst and materials development.

The most used CSC installed software is Gromacs

Molecular dynamics is one of the most used simulation techniques on CSC supercomputers. To facilitate their use, CSC installed several open source MD applications and acquired a license for several commercial ones. In fact, the most used CSC installed application is Gromacs, which is particularly efficient and versatile open source software. Especially, Gromacs is a highly scalable software in the Mahti supercomputer.

The most important qualities of MD software are its ease of use, accuracy of the available force fields, available methods and analysis tools and of course its efficiency i.e. how big systems can be studied in a reasonable time and with reasonable electricity consumption. Big model systems require using thousands of CPU cores to work on the same simulation problem. CSC’s supercomputers are very well suited for these due to their efficient CPUs and very fast interconnect. In an accurate simulation, the positions of all atoms are needed to calculate the forces and this means a lot of communication between the CPUs, which in turn requires a very fast interconnect network.

Another important trend in both supercomputing and molecular dynamics simulations is the use of GPUs i.e. Graphics Processing Units to speed up the calculations. Several of the MD simulation codes available at CSC are capable of utilizing GPUs which further makes the simulations efficient. This is particularly important, since LUMI, the 3rd fastest supercomputer in the world, increases the GPU capacity at the Kajaani datacenter significantly and that capacity is thus available to researchers using these methods.

The mere existence of clever software and abundant computing capacity is of course not enough. Significant improvements in the usage efficiency can be reached by using the software and the hardware optimally. In the past few years CSC has collaborated with BioExcel to organize introductory and advanced hands-on courses to make the researchers aware of the latest developments and best practices on using Gromacs and related automation and scripting tools. Employing the latest algorithms, using optimized code and automating save energy, use the resources more efficiently and enable more science to be done with the existing hardware.

Use cases, a few examples

Professor Hanna Vehkamäki and her group (University of Helsinki) use MD to study collisions between atmospheric molecules, ions and clusters. These collisions can lead to the formation of stable clusters and they can continue to grow through additional sticking collisions into aerosol particles. Aerosol particles in turn affect our climate, by scattering solar radiation and functioning as nuclei for cloud formation, and health, by inhalation into the lungs. The research group performs molecular dynamics collision simulations and potential of mean force calculations of atmospherically relevant molecules, ions, and clusters to obtain a robust prediction of their collision rate coefficient. Their aim is to derive improved theoretical models and parametrizations to predict collision rates between atmospheric molecules, ions, and clusters.

In another study, Vehkamäki’s group is using Gromacs to study the hidden role of gases in trees. Water in tree tissue is held under negative pressure and this causes air to enter from the roots in the form of nanobubbles coated with lipids. In this study the goal is to understand nanobubble stability as the pressure becomes more negative, and to extend the so called ‘cohesion tension theory’ to the molecular level.

With his research group, Professor Ilpo Vattulainen (University of Helsinki) has plenty of projects utilizing MD methods. We give here two examples.

Vattulainen et al. are studying the dimerization of main protease of SARS-CoV-2 virus. This protease, that the SARS-CoV-2 researchers call as Mpro, is required for maturation of the virus, a process which allows these viruses to exit infected cells and begin another cycle of further infections. By blocking the activity of Mpro with drugs, the virus can be inactivated. However, the mechanism how this main protease functions has remained unclear.

The researchers explored various stages of Mpro activation with all-atom molecular dynamics simulations. Using machine learning (ML)k models derived from the simulation data, they studied the dynamics of Mpro, resolved mechanisms that regulate the activity Mpro, and unveiled how the virus regulates the timing of its maturation. They found compelling evidence that dimerization of two Mpro monomers regulates the activity of the protease. The results suggest an alternative strategy to vaccination, where previously discovered antiviral drugs can be repurposed to successfully combat the spread of Covid-19 infections by blocking the dimerization of Mpro.

In another study Vattulainen and Biological Physics Group at University of Helsinki studied how antidepressant drugs bind. Their biomolecular simulations combined with experimental studies showed that antidepressants directly bind to Tyrosine kinase receptor 2 (TrkB) acting as a membrane receptor for the brain-derived neurotrophic factor (BDNF). A fascinating feature in this binding process is the central role of cholesterol. The study revealed that TrkB senses membrane cholesterol, which modulates the conformation of TrkB and thus regulates the ability of antidepressants to bind to the receptor.

Professor Gerrit Groenhof at University of Jyväskylä is interested in the interplay between light and matter and he is researching organic photovoltaics that hold great promise for a sustainable future, but their applicability is severely limited by a relatively short exciton diffusion length within the material.

Groenhof et al. are using massively parallel computing to explore in atomic detail if this limitation can be exceeded by strongly coupling the material to confined light modes, as occur for example in Fabry-Pérot optical cavities. They performed multi-scale molecular dynamics simulations of over 1000 photoactive molecules, strongly coupled to the confined light modes of Fabry-Pérot cavity. The simulations with the GROMACS and Gaussian16 programs on the Mahti supercomputer suggest that under strong light-matter coupling exciton transport can be enhanced by orders of magnitude, reaching distances over ten micrometers. This is significant in the context of improving photovoltaics applicability.

Atte Sillanpää
Author works in customer support including chemistry and international projects.

Tommi Kutilainen
The author has a few decades of experience in scicomm at CSC.
Twitter: @TommiKutilainen