Development of hybrid coarse-grain/atomistic simulation models and their application to membrane-bound proteins

"Hybrid simulation methods have been developed to explore biological membranes and the behaviour of drugs and proteins therein."

archer logo

Want HECtime on ARCHER?

Interested in how to apply for time on ARCHER like these researchers did?

You can find out more here

The membrane environment is crucial for the efficient functioning of living systems. It acts as a semi-permeable barrier maintaining the integrity of cells, and many important biological process arise from the proteins bound within the membrane. Computer simulations of membrane systems allow us to understand their structure and function – the membrane environment affects the bound proteins, while drugs must pass through membranes to reach their active site.

However, in conventional molecular dynamics simulations all the atoms are represented, making these calculations very time consuming. A solution to this problem is to simplify the representation of the system by merging groups of atoms into larger beads. Such a coarse-grain approach, while faster, is less accurate.

To maintain the overall accuracy of our calculations, while improving efficiency, we have therefore developed a hybrid model in which the membrane molecules and surrounding solvent are represented at a coarse-grain level, while bound proteins and drugs are modelled at the atomistic level – we use the higher resolution model for the more important parts of the system.


To develop and test such a model, we calculated how the amino-acid side chains pass through the membrane, and how a range of peptides tilt and fold in the membrane (figures 1 and 2). Our model was able to reproduce the results seen using more complex atomistic models. 

case study 3 figure 1

Figure 1 - Schematic showing our hybrid simulation model. Reprinted with permission from Copyright 2017 American Chemical Society.

case study 3 figure 2

Figure 2 - Snapshots from hybrid simulations of Kalp23 and Walp23 peptides in either a DOPC or DMPC membrane. Water beads are shown in blue, lipids in yellow, and the AA helix in purple cartoon representation. Reprinted with permission from Copyright 2017 American Chemical Society.

case study 3 figure 3

Figure 3 - Experimental versus predicted log D and b boxplot of absolute deviations compared to experiments. Reprinted with permission from Copyright 2017 American Chemical Society.

We have subsequently tested the model in a blind trial to predict the distribution coefficient of small molecules between water and cyclohexane. The distribution coefficient reflects how readily the small molecule separates into the aqueous and non-aqueous phases. In our approach the water and cyclohexane solvents are modelled at a coarse-grain level, and the small molecules at the atomistic level. In the blind trial, we were able to predict the distribution coefficients with accuracy comparable to that of more detailed atomistic models (figure 3).

Key Outputs

Methodology has been implemented and is now available in one of the major simulation programs, LAMMPS. This means that it is available for anyone to use.

The ability to simulate proteins and drugs in the membrane environment.

Allowed training of staff.


Key Academics


University of Southampton - UK

Prof. Jonathan Essex


University of Gothenburg - Sweden

Dr Samuel Genheden



For more information about this study, please contact Jonathan Essex - This email address is being protected from spambots. You need JavaScript enabled to view it.


Grant Information

This study made use of a HECBioSim ARCHER project allocation EP/L000253/1 and was sponsored by the Wenner-Gren foundation.

southampton gothenburg