9.6 Analysis of PDB Membrane Files with APL@Voro


APL@Voro can be started by using the File menu entry:

This menu entry will only be shown if it was activated in the 4.2 Web and Local Connections. To install APL@Voro on your local system, please download it from http://www.aplvoro.org. This program was developed for unix-compatible systems (also mac) and there is a number of pre-compiled download versions available.

Please be sure to install APL@Voro version >= 2.9, because this is the first version providing full PDB import functionality by command line.

Why APL@Voro?

The idea to create the tool APL@Voro goes back to a drawback of the MembraneEditor. The following screenshot shows the way how the MembraneEditor presents the area per lipid:

In this example, the area per lipid is stated with 32.73 angstrom for the extracellular side and 33.17 angstrom for the intracellular one. This is of course only a very rough computation by computing the complete membrane area and subdividing it by the number of total lipids on one membrane side.

Obviously, this approach might be appropriate if the value is computed for a homogeneous membrane containing only a single lipid type. But already for two different lipid types, e.g. Cholesterol and a Phospholipid, this formula is too imprecise. Moreover, it is not possible to analyze the area of a specific lipid, only a global view is possible.

The solution to these problems is now provided by Gunther Lukat: APL@Voro; Area Per Lipid computation by using Voronoi diagrams.

To find out more about the theory behind this tool, please read the following publication:

Lukat G, Kr├╝ger J, Sommer B: APL@ Voro: A Voronoi-Based Membrane Analysis Tool for GROMACS Trajectories (2013). J. Chem. Inf. Model. 53(11), 2908-2925.

On this help page some basic information about the features of APL@Voro and the use of APL@Voro in conjunction with the MembraneEditor will be provided. There is also a PDF documentation available at the original APL@Voro website:


MembraneEditor Membranes in APL@Voro

The following image shows a membrane generated with the MembraneEditor, exported to PDB format and then imported into APL@Voro. This is the extracellular side of the membrane. The typical Voronoi cells are shown on the right side, and on the left side: the list of lipids. Obviously, there are six different lipid types.

The Voronoi cells were computed based on key atoms. More information about this topic is found below.

Examining now the original membrane in the MembraneEditor, the left side of the image - the Overview window - shows also six lipid types. But if we compare both images, we can see that a) the different lipid types are not colored and b) that the microdomain in the center of the membrane is not visible in the Voronoi representation of APL@Voro.

First, let us change the colors in APL@Voro. By right-clicking on the lipid type in the left window, the colors can be changed. Here, a color scheme is used compatible to the one of the MembraneEditor:

But now we also want to make the raft - located in the center of our membrane - visible. For this purpose, we have to export a new PDB file from the MembraneEditor. But this time, we will make use of the option Write DomainID in ResName found in the PDB properties:

In the following example, the DomainID is written at the 2nd position and the LayerID is not used at all. Please have a look at 9.1 Pdb output settings - Dealing with limitation and compatibility issues to learn more about this option.

Now, if you click in MembraneEditor File -> Analyze Membrane with APL@Voro again (there is also a nice button in the toolbar for fast access), a new membrane is generated, but this time it contains the microdomain. The lipid entries on the left side show, how it works: the ResNames are combined with the ID of the microdomains and listed as separate entries. Now, the coloring scheme of the MembraneEditor can also be manually applied to the lipids, the result is shown here:

Now, there are different ways to analyze and visualize the area per lipid. Here, the area per lipid is shown as a three-dimensional representation. This representation can also be used to visualize the membrane thickness.

Of course, to evaluate the area per lipid precisely, the concrete values have to be examined. Now, we want to examine the area per lipid for all lipids contained in the microdomain. For this purpose, we only have to select all lipid types which are part of the microdomain (here indicated by the DomainID 1):

Now, the menu entry Windows -> Show/hide the Averages Window has to be activated. There are two sections in the appearing window. All shows the values for all lipids. The name of the key atom, the position, the area per lipid as well as the thickness can be found there:

Changing now to the Selected section, only those lipid types will be shown which were selected. Because only the extracellular side (or Upperside) was selected, the following image does not show any values for the intracellular side:

To work also with the intracellular side, just toggle Windows -> Show/Hide the Downside Voronoi view.

APL@Voro provides a number of different approaches to visualize the area per lipid and the membrane thickness. E.g. it is also possible to generate 2D plots and to compare it with the other visualization options. Moreover, proteins can also be taken into account, but for this purpose a index.ndx file is needed, see below. Moreover, complete Molecular Dynamics simulations performed with Gromacs can be evaluated with APL@Voro.

Please see the documentation at http://www.aplvoro.org and/or the publication stated above for more details.

How does the communication work?

If APL@Voro is started separately, the user has to define the key atoms manually. There are different key atom selection schemes available and it depends on the application case which is favorable for a particular purpose. The APL@Voro publication stated above lists some of these options. The MembraneEditor passes P atoms - e.g. in case of Phospholipids - or O atoms - e.g. in case of Cholesterol - to APL@Voro. This simple selection model defines the headgroups of the lipids as the starting point for the Voronoi cell generation.

If the user wants to change this configuration, there are two options to change the settings:

  • First, it is possible to reload the temporary PDB file generated by the MembraneEditor into APL@Voro and define all values manually. The file is located in your home/Cm2/temp folder.
  • Second, the user can generate the .aplvoro config file with the MembraneEditor and then edit it manually. The .aplvoro config file export has to be activated in the PDB Properties dialog, see chapter 9.1 Pdb output settings - Dealing with limitation and compatibility issues.

MembraneEditor generates the .aplvoro config file in the background and passes it together with the PDB file to APL@Voro by using the command line:

aplvoro -f file.pdb -c file.aplvoro

The box size is stored in the PDB file in the CRYST entry by MembraneEditor and this way directly transmitted to APL@Voro.

This .aplvoro config file can be edited by the user. More information are found at the documentation at http://www.aplvoro.org.

Protein Handling and Workarounds

The handling of proteins is not directly supported by the communication between APL@Voro and MembraneEditor, because an .ndx file is required. This file can be generated by Gromacs. If a protein-holding membrane is generated by the MembraneEditor and directly exported to APL@Voro, the protein will be ignored. Usually, this will lead to artifacts: the lipids surrounding the protein will have a very high area per lipid because their Voronoi cells are merged with the missing areas of the protein (residues).

For a first analysis, this is usually not a problem. Especially, the user can a) select all lipids and b) remove only those lipids from the selection which are in direct neighborhood to the protein and then evaluate only the selected areas and their averages. This will lead to a quite precise estimation of the area per lipid.

But if the area occupied by the protein's membrane domains should be correctly computed, an .ndx file has to be generated. Check the Gromacs manual for this purpose.