This news blog provides news about the e-IRG and related e-Infrastructure topics.

Back

Computational microscopy needs big supercomputers to study secrets of self-organising cell membranes

At the occasion of the Super D SURFsara event, we had the opportunity to talk with Siewert-Jan Marrink, Professor of Molecular Dynamics at the University of Groningen. The work Professor Marrink does in the Molecular Dynamics Group is to use big computers to simulate the interaction of molecules, and in particular, the molecules that build cell membranes. He and his colleagues are interested in understanding how cellular membranes are organised, how the lipids and the proteins that make up these membranes are self-organised. This is a very complex matter because cell membranes are composed of hundreds of different lipid type of molecules crowded with proteins. They together make or give rise to the different processes that govern the functioning of cells.

To the question why cell membranes are so complex Professor Marrink answered that if you think about the basic functionality of a cell membrane, it is just to provide a barrier between the inside and the outside of the cell. For that, you only need lipid molecules that form a by-layer. Any lipid molecule can do that. Still, there are hundreds of different types of lipids and one is only at the beginning of trying to understand why there is a certain need for such a large diversity. It seems that many lipids specifically bind to proteins and are able to bring proteins together - and not just any proteins but proteins that also need to function together. The lipids seem to have a nice steering role in governing the organisation of the cell membrane.

In the old days you would take a cell, prepare it, put it under a microscope and look at the structure. That is not the way that you do this today, we wanted to know.

Professor Marrink explained that you can still do this. Many experimental colleagues look at membranes in that way but the main problem is that the resolution you can attain with the traditional microscope techniques is very limited. You can not see the individual lipids and proteins that are the basic building blocks of the membrane. Professor Marrink's group performs computational microscopy where one actually simulates the interaction between all the components in a virtual reality. The advantages that you have with reaching a resolution at the atomic level of the molecules and at time scales that go from the picoseconds to the nanoseconds and microseconds is that you can resolve the dynamics. This is the big advantage of the computational microscopy approach over traditional experimental microscopy.

We summarized that you build the model, put it in the computer and perform the calculations.

Professor Marrink confirmed that his group solves the equations of motions that were already formulated by Newton many centuries ago. The atoms exert forces on each other. They make active rise to the molecules so that they start moving around. The researchers can simulate this if they have enough computational resources in order to visualize what will happen to the system over a period of time. This is usually microseconds to up to milliseconds of time that they can simulate.

As for the computational resources, this is not something you can do at home. It is only a very small and limited part of the cell membrane that you could simulate on a laptop at home. If you really want to go to the larger systems and follow the dynamics for an interesting amount of time, you need the biggest supercomputer that you can basically lay your hands on.

We wanted to know why the model requires so much computing time.

Professor Marrink said that you have to compute the forces between all the particles that are in the system. This can be millions and millions of atoms. The researchers solve this iteratively so they want one time step at a time. The time steps they can take between different snapshots of the system are of the order of a couple of femtoseconds. In order to get to microseconds they have to have millions and millions of these sequential time steps that they have to calculate. That is a huge computational effort.

Currently, Professor Marrink's team is using the supercomputers at SURFsara, we asked whether he is also using other supercomputers.

Professor Marrink said that the researchers have their own local supercomputers at the University of Groningen but at the national level, they rely on the SURFsara supercomputers. Sometimes, the researchers are also involved in European projects but the bulk of their simulation is done in Amsterdam.

At the European scale one is working on an exascale supercomputer by 2022, we asked if this would be helpful to solve the simulation challenges.

Professor Marrink answered that he wouldn't say that this would solve the problems but it would definitely help. Earlier on, the researchers were always facing the limit of the computational resources that they have. If they really can get to the exascale level, this would open up a whole new box of simulations and systems that they can look at and processes that they can study because still at this point, they are limited in time scales that are far away from a lot of cellular processes that take place on the time scales of seconds or more that they currently can not reach with the current computational power.

Sometimes, you would like to simulate the model of a complete cell. When would that be possible, we wanted to know.

That is the goal for in about ten years' time. Professor Marrink hopes that by then his team can construct the model of a full biological cell, including the cell membrane with all the different lipids and proteins that are found there but also the interior of the cell, the cytoplasm and some of the internal organelles in order to get to a complete cell model where you still have access to the molecular resolution that gives rise to the whole functioning of the cell. He thinks that this is required because all the components are interconnected with each other. Of course, you can study processes in isolation but at the end the cell functions because all the processes are connected. This, eventually, his team hopes to also capture in their simulation model.

We are looking forward to the presentation in ten years' time.

Professor Marrink smiled that this was his promise to show a nice movie of a complete cell simulation.