MDfit

Molecular Dynamics fitting (MDfit) :

Molecular dynamics cryo-EM fitting on a desktop.

Download MDfit:

Download from UCSD mirror.

Mirror for the website:

MDfit UCSD mirror.

Structure-based force field (SMOG):

SMOG

SMOG is the force field upon which MDfit fit is based. No installation of SMOG is required to run MDfit. Below we guide you to instructions for generating SMOG force fields.

Sanbonmatsu Team:

Sanbonmatsu Team.

Onuchic Group:

Onuchic Group.

Description of MDfit Approach

Molecular complexes often visit particular configurations only transiently. For these systems, direct structural methods (such as x-ray crystallogaphy and cryogenic-electron microscopy) can not always provide every atomic detail since these configurations do not correspond to minima on the energy landscape of that molecule. Nonetheless, many of the structural features of these transient configurations may be consistent with information which is accessible from x-ray/cryo-em studies. Additionally, additional experimental methods (such as chemical probing) may complement this data and provide partial information about the structural details of these transient configurations.

MDfit is a computational methodology that allows you to incorporate information from x-ray crystallography, cryo-em and biochemical studies in order to prepare atomic models for these configurations. The MDfit method starts with a structure-based model (such as those provided by the smog@ctbp server). In a structure-based model, the initial configuration (obtained from x-ray/cryo-em, or other methods) is defined as the lowest energy configuration (red line in figure). For a description of the all-atom structure-based potentials, check out our articles in Proteins and Biophysical Journal. If a simulation is performed with the structure-based forcefield alone (at low enough temperatures), it will provide you with a description of the fluctuations about the initial basin. Next, if cryo-em data is available for an alternate configuration of your system (or a sub-region of your complex), you can add another term to the forcefield that is based on the correlation between your system (or sub-region of it) and the cryo-em density (green line in figure). If you include the cryo-em term of the forcefield with a strong enough energetic weight, then the total potential energy surface will be "downhill" (blue line) with the target (cryo-em) configuration at the bottom of the new energetic minimum. Since the target is a minimum on the energy landscape, performing simulations with this new forcefield will result in the system moving into the target configuration. When you have additional information about your target system (for example, regions that lack an em density), then restraints may be introduced via a variety of types of interactions (i.e. harmonic, 6-12, etc). Again, when simulations are performed with the combined structure-based/em-restraint forcefield, the system can relax into configurations that are consistent with all of these contributions.

Downloading the code

We are in the process of preparing a "git" version of MDfit updates for Gromacs so that the newest version of gromacs may be used with MDfit, though it is not ready yet. The code/tutorial should be considered a Beta version. We are happy to receive feedback on it, but use for production runs at your own risk...

Currently, this method is available in a modified version of the open source molecular dynamics package Gromacs (Version 4.0.5). The beta version of the MDfit source code, which exists as a modified version of gromacs can be downloaded here. We are in the process of preparing a "git" version of MDfit updates for Gromacs so that the newest version of gromacs may be used with MDfit, though it is not ready yet.
Notes on compiling and using this code:

1. You must compile with MPI support and perform simulations on a minimum of 2 processors. This is not usually an issue since systems studied by cryo-em are typically large (and they scale to many processors). Additional notes on compiling can be found here.
2. The simulated map calculation is performed on 1 processor. This can lead to the map calculations being rate limiting. For this reason, you don't need to calculate the map at every step. Depending on the strength of the map, you can update the map forces every several hundred timesteps.
3. This code has been tested on systems running MAC OSX 10.6 (64 bit Intel) and several Linux distributions. We appreciate feedback, if you have trouble compiling or running the code (just email smog@ctbp.ucsd.edu). We suggest working through the example below with your installed version, to ensure the code it operating properly.

The code/tutorial should be considered a Beta version. We are happy to receive feedback on it, but use for production runs at your own risk...

After installing the code on your local machine, follow these steps to perform a modeling simulation with MDfit. Here, we provide an example (For Adenylate Kinase) of how to use the code. Adenylate Kinase is a 3-domain protein that undergoes a large structural rearrangement upon ligand binding. In this example, we will start with the open conformation of AKE (PDB: 4AKE), and we will fit it to a theoretical density generated from the closed conformation (PDB: 1AKE). You can download the tarball with all files that are referenced in this tutorial, or you can download them individually as you proceed.

1. Download a PDB file (or, use any other PDB-formatted structure) For this example, PDB entry 4AKE is used (it is also in the tarball). In this PDB file there are 2 copies of AKE. We will only need 1 copy for this exercise, so remove 1 copy from the file (4AKE.single.pdb).
2. Prepare your em density For this exercise, use this density (1AKE.density.sit). Note: If you are using periodic boundary conditions, make sure the target density has positive coordinates. If you do not have a situs file (.sit), you will need to reformat it. Often, one has a .brix file. To obtain a situs file, we load the brix file into Chimera. Then, go to "Volume->Volume Viewer". Inside of the Volume Viewer window, go to "File->Save map as...". Select your new file name and select File type MRC. Then, using the map2map module of Situs, convert the mrc file to a Situs file. There may be simpler ways to convert a brix file to a Situs file, but this will get you started. This file also has 1 additional field in the first line. It is a real number that represents the threshold value TR for the density. Voxels that have values smaller than TR will be given the value TR. In the sample situs file, this value is 0.0. If you don't want to use a threshold, then provide a number that is lower than any voxel value in your .sit file.
3. Perform an initial rigid-body alignment of the atomic model to the target density This can be performed manually, as done here, or you can use automated alignment methods. A manually-aligned structure can be obtained here (4AKE.aligned.pdb).
4. Generate a topology and coordinate file for your system, using the SMOG webtool. Here are output topology (4AKE.aligned.top) and coordinate (4AKE.aligned.gro) files generated by the SMOG server, with default values.
5. Correct any shift in coordinates that may have occurred The SMOG server often translates your system. This can be corrected for using the trjconv module of Gromacs. First, determine if the coordinates have shifted, and by how much. To calculate the shift use any atom in the input pdb (4AKE.aligned.pdb) and output coordinate file (4AKE.aligned.gro) as reference. For the example, the following command will correct for the shift (select "system" for the output). If you use periodic boundary conditions, make sure the box (defined in the last line of the .gro file) is large enough that your molecule will not hit the boundary and then "jump" away from the map. For this case, we added 10 nm to each dimension of the box (manually).

   trjconv -f 4AKE.aligned.gro -s 4AKE.aligned.gro -trans 3.9532 3.5076 3.736 -o 4AKE.corrected.gro

6. Prepare your tpr file Download a sample mdp file (em.MDfit.mdp). Take a careful look at this sample mdp file, as there are a few necessary settings that are specific to MDfit. Now, you have all the necessary files to perform your first fit. As with a standard release of gromacs, the grompp module will be used to generate a .tpr file. Here is a sample grompp command:

   grompp -f em.MDfit.mdp -c 4AKE.corrected.gro -p 4AKE.aligned.top -o 4AKE.MDfit.tpr

7. Perform the fit With the MDfit-modified gromacs, perform the simulation:

   mpirun -np 2 ~/BIN/gromacs-em-beta/bin/mdrun -s 4AKE.MDfit.tpr -mmff -emf 1AKE.density.sit -v

   On most modern machines, this calculation should only take a few minutes.

8. Check the results You can view the fitting simulation using VMD. You can check the quality of the fit, for this example, by comparing it to the closed conformation of AKE (1AKE). If you perform an rmsd alignment of the C-alpha atoms, the final configurations of the fit should be ~ 1.5 Angstroms from the closed configuration. Your results should look like the following figure.

In addition to incorporating data from electron density maps, there are many ways in which you might want to incorporate additional information during your runs. For the following descriptions, you can use any version of Gromacs, if you are not including em information. If you want to use special restraints, in addition to the em information, then use the version of the code from this page.

Here we will provide several examples of how you might add additional restraints into your system. Since the structure-based forcefields are very simple, it is straightforward to include additional interactions into your topology file. We will also provide a variety of helpful scripts to get you started.

References

MDfit

Ratje et al. (2010) "Head swivel on the ribosome facilitates translocation by means of intra-subunit tRNA hybrid sites" Nature 468, 713-716.

Webserver

Noel JK, Whitford PC, Sanbonmatsu KY & Onuchic JN (2010) "SMOG@ctbp: simplified deployment of structure-based models in GROMACS" Nucleic Acids Res. DOI: 10.1093/nar/gkq498.

All-atom Protein Model

Whitford PC, Noel JK, Gosavi S, Schug A, Sanbonmatsu KY & Onuchic JN (2009) "An all-atom structure-based potential for proteins: Bridging minimal models with all-atom empirical forcefields." PROTEINS DOI: 10.1002/prot.22253.

All-atom RNA Model

Whitford PC, Schug A, Saunders J, Hennelly SP, Onuchic JN & Sanbonmatsu KY (2009) "Non-local helix formation is key to understanding S-adenosylmethionine-1 riboswitch function." Biophys. J. 96, L7-9. DOI: 10.1016/j.bpj.2008.10.033.

C-alpha Protein Model

Clementi C, Nymeyer H & Onuchic JN (2000) "Topological and energetic factors: What determines the structural details of the transition state ensemble and En-route intermediates for protein folding? An Investigation for small globular proteins." J. Mol. Biol. 298, 937-953. DOI:10.1006/jmbi.2000.3693

Please direct questions and comments to smog@ctbp.ucsd.edu

Page created and maintained by Karissa Sanbonmatsu (kys@lanl.gov) and derived from page by Whitford, Paul (p.whitford@neu.edu) and Jeff Noel (jknoel@gmail.com).