This page is adopted from a page by Bert de Groot and modified to be compatible with Gromacs 5.x
- Preparation of a typical MD simulation
- Setup the simulation system
- Carry out the simulation
- Trajectory analysis
- Optional exercises
In the previous part we’ve learned what MD simulations are and how to simulate a van der Waals gas. Now it is time to set up a simulation of a
biological macromolecule: a small protein.
Proteins are nature’s universal machines. For example, they are used as building blocks (e.g. collagen in skin, bones and teeth), transporters
(e.g. hemoglobin as oxygen transporter in the blood), as reaction catalysts (enzymes like lysozyme that catalyse the breakdown of sugars), and as nano-machines (like myosin that is at the basis of muscle contraction). The protein’s structure or molecular architecture is sufficient for some of these functions (like for example in the case of collagen), but for most others the function is intimately linked to internal dynamics. In these cases, evolution has optimised and fine-tuned the protein to exhibit exactly that type of dynamics that is essential for its function. Therefore, if we want to understand protein function, we often first need to understand its dynamics (see references below).
Unfortunately, there are no experimental techniques available to study protein dynamics at the atomic resolution at the physiologically relevant time resolution (that can range from seconds or milliseconds down to nanoseconds or even picoseconds). Therefore, computer simulations are employed to numerically simulate protein dynamics.
As before, we will use the GROMACS simulation package for this.
Today, we will simulate the dynamics of a small, typical protein domain: the B1 domain of protein G. B1 is one of the domains of protein G, a member of an important class of proteins, which form IgG binding receptors on the surface of certain Staphylococcal and Streptococcal strains. These proteins allow the pathogenic bacterium to evade the host immune response by coating the invading bacteria with host antibodies, thereby contributing significantly to the pathogenicity of these bacteria (click here for further background information on this protein). We will now follow a standard protocol to run a typical MD simulation of a protein in a box of water in gromacs. The individual steps are summarized in a flowchart on the right site
Before a simulation can be started, an initial structure of the protein
is required. Fortunately, the structure of the B1 domain of protein G has been solved experimentally, both by x-ray crystallography and NMR. Experimentally solved protein structures are collected and distributed by the Protein Data Bank (PDB). Please open this link in a new browser window and enter “protein G B1” in the search field. Several entries in the PDB should match this query. We will choose the x-ray structure with a high resolution (entry 1PGB with resolution of 1.92 ang) for this study. To download the structure, click on the link “1PGB”, and then, under “Download Files”, select “PDB File”. When prompted, select “save to disk”, and save the file to the local hard disk (click here if that does not work). To have a look at the contents of the file, on the unix prompt, type:
As we’ll learn in the next practical on protein structure, the file starts with general information about the protein, about the structure, and about the experimental techniques used to determine the structure, as well as literature references where the structure is described in detail. (in “more”, press the spacebar to scroll). The data file contains the atomic coordinates of our protein structure with one line per atom. (quit the “more” program by pressing “q”). Now we can have a look at the structure:
to visualise the structure. Try using “Graphics -> Representations -> Drawing method” and set it to “New Cartoon” or “Licorice”. End vmd using “exit”at the console prompt.
Question: Why do we start our MD simulations from the experimentally determined 3D structure? Isn’t it enough to know the proteins amino acid sequence?
We will now prepare the protein structure to be simulated in gromacs. Although we now have a starting structure for our protein, one might have noticed that hydrogen atoms (which would appear white) are still missing from the structure. This is because hydrogen atoms contain too few electrons to be observed by x-ray crystallography at moderate resolutions. Also, gromacs requires a molecular description (or topology) of the molecules to be simulated before we can start, containing information on e.g. which atoms are covalently bonded and other physical information. Both the generation of hydrogen atoms and writing of the topology can be done with the gromacs program pdb2gmx:
if needed type: source /usr/local/gromacs/bin/GMXRC
gmx pdb2gmx -f 1PGB.pdb -o conf.pdb
when prompted for the force-field to be used, choose the number corresponding to the OPLS-AA/L all-atom force field, and SPC/E for water. View the result with:
See the added hydrogens in the “Licorice” representation? The topology file written by pdb2gmx is called “topol.top”. Have a look at the contents of the file using:
you will see a list of all the atoms (with masses, charges), followed by bonds (covalent bonds connecting the atoms), angles, dihedral angles etc. Near the very end of the topology (in the “[molecules]” section) there is a summary of the simulation system, including the protein and 24 crystallographic water molecules.
The topology file thus contains all the physical information about all interactions between the atoms of the protein (bonds, angles, torsion angles, Lennard-Jones interactions and electrostatic interactions).
The next step in setting up the simulation system is to solvate the protein in a water box, to mimick a physiological environment. For that, we first need to define a simulation box. In this case we will generate a rectangular box with the box-edges at least 7 Angstroms apart from the protein surface:
gmx editconf -f conf.pdb -o box.pdb -d 0.7
(note that gromacs uses units of nanometers). View the result with
and, in vmd , use:
Extension –> Tk Console —> type “pbc box_draw”
Now, exit vmd and fill the simulation box with SPC water using solvate:
gmx solvate -cp box.pdb -cs spc216 -o water.pdb -p topol.top
Again, view the output (water.pdb) with vmd
Now the simulation system is almost ready. Before we can start the dynamics, we must perform an energy minimisation.
Question: Why do we need an energy minimisation step? Wouldn’t the crystal structure as it is be a good starting point for MD as it is?
For the energy minimisation, we need a parameter file, specifying which type of minimisation should be carried out, the number of steps, etc. For your convenience a file called “em.mdp” has already been prepared and can be downloaded from here. View the file with “more” to see its contents. We use the gromacs preprocessor to prepare our energy minimisation:
gmx grompp -f em.mdp -c water.pdb -p topol.top -o em.tpr -maxwarn 2
This collects all the information from em.mdp, the coordinates from water.pdb and the topology from topol.top, checks if the contents are consistent and writes a unified output file: em.tpr, which will be used to carry out the minimisation:
gmx mdrun -v -s em.tpr -c em.pdb
The output shows that already the initial energy was rather low, so in this case there were hardly any bad contacts. Having a look at “em.pdb” shows that the structure hardly changed during minimisation.
The careful user may have noticed that grompp gave a warning:
System has non-zero total charge: -4.
Before we continue with the dynamics, we should neutralize this net charge of the simulation system. This is to prevent artifacts that would arise as a side effect caused by the periodic boundary conditions used in the simulation. A net charge would result in an electrostatic repulsion between neighboring periodic images. Therefore, 4 sodium ions will be added to the system:
gmx genion -s em.tpr -o ions.pdb -np 4 -p topol.top
Select the water group (“SOL”), and 4 water molecules will be replaced by sodium ions. The output (ions.pdb) can be checked with vmd. To better see the ions, use vdW representation for the ions:
Just to be on the safe side, we repeat the energy minimisation, now with the ions included (remember to (re)run gmx grompp to create a new run input file whenever changes to the topology, or coordinates have been made):
gmx grompp -f em.mdp -c ions.pdb -p topol.top -o em.tpr -maxwarn 2
gmx mdrun -v -s em.tpr -c em.pdb
Now we have all that is required to start the dynamics. Again, a parameter file has been prepared for the simulation, and can be downloaded here. Please browse through the file “md.mdp” (using “more”) to get an idea of the simulation parameters. The gromacs online manual describes all parameters in detail here. Please don’t worry in this stage about all individual parameters, we’ve chosen common values typical for protein simulations. Again, we use the gromacs preprocessor to prepare the simulation:
gmx grompp -f md.mdp -c em.pdb -p topol.top -o md.tpr -maxwarn 2
and start the simulation!
gmx mdrun -v -s md.tpr -c md.pdb -nice 0
The simulation is running now, and depending on the speed and load of the computer, the simulation will run for a number of minutes.
How do the parts of energy minimization and MD simulation differ (with reference to energy landscapes)?
D. Analysis of gromacs simulations
The simulation is running now (or finished) and we can start analyzing the results. Let us first see which kind of files have been written by the simulation (mdrun):
We see the following files:
- traj.xtc – the trajectory to be used for analyses
- traj.trr – the trajectory to be used for a restart in case of a crash
- ener.edr – energies
- md.log – a LOG file of mdrun
- md.gro – the final coordinates of the simulation
vmd em.pdb traj_comp.xtc
For a quantitative analysis on the protein fluctuations, we can view how fast and how far the protein deviates from the starting (experimental) structure:
gmx rms -s md.tpr -f traj_comp.xtc
When prompted for groups to be analyzed, type “1 1”. gmx rms has written a file called “rmsd.xvg”, which can be viewed with:
We see the Root Mean Square Deviation (rmsd) from the starting structure, averaged over all protein atoms, as a function of time.
Question: Why is there an intial rise in the rmsd?
If we now wish to see if the fluctuations are equally distributed over the protein, or if some residues are more flexible than others, we can type:
gmx rmsf -s md.tpr -f traj_comp.xtc -oq -res
Select group “3” (C-alpha). The result can be viewed with:
We can see that mainly four regions in the protein show a large flexibility: around residues 1, 11, 21 and 38. To see where these residues are located in the protein, type:
Under “Colors”, select “beta”. The protein backbone is now shown with the flexibility encoded in the color. The red regions are relatively rigid and the blue regions are relatively flexible. It can be seen that the alpha-helix and beta-sheet are relatively stable, whereas the loops are more flexible.
The simulation not only yields information on the structural properties of the simulation, but also on the energetics. With the program gmx_energy the energies written by mdrun can be analysed:
gmx energy -f ener.edr
Select “Potential” and end your selection by pressing enter twice, View the result with:
As can be seen, the total potential energy initially rises rapidly after which it relaxes again.
Question: Can you think of an explanation for this behavior?
Please repeat the energy analysis for a number of different energy terms to obtain an impression of their behavior.
Question: Do you think the length of our simulation is sufficient to provide a faithful picture of the protein’s conformations at equilibrium.
We continue with a number of more specific analysis, the first of which is an analysis of the secondary structure (alpha-helix, beta-sheet) of the protein during the simulation.
The next thing to analyze is the change in the overall size (or gyration radius) of the protein:
gmx gyrate -s md.tpr -f traj_comp.xtc
(again, select group “1” for the protein)
The analysis shows that the gyration radius fluctuates around a stable value and does not show any significant drift. Another important check concerns the behaviour of the protein surface:
gmx sasa -s md.tpr -f traj_comp.xtc
(again, select group “1” for surface calculation and group “1” for output)
Now view the total solvent accessible surface area with:
xmgrace -nxy area.xvg
We now see three curves together: black for the hydrophobic accessible surface, red for the hydrophilic accessible surface, and green for the total accessible surface.
Question: Is the total
(solvent-accessible) surface constant? Are any hydrophobic groups exposed during the simulation?
In a next step we perform a secondary structure calculation using the DSSP program
gmx do_dssp -s md.tpr -f traj_comp.xtc -ver 1
gmx xpm2ps -f ss.xpm
and view the result with:
if “gv” is not installed on your computer, try “xpsview”, “ghostview” or “gs”. As can be seen, the secondary structure is rather stable during the simulation, which is an important validation check of the simulation procedure (and force-field) used.
- You’ve probably noticed that in the simulation about only ten percent of the system that was simulated consisted of protein, the rest was water. As we are mainly interested in the protein’s motions and not so much in the surrounding water, one could ask if we couldn’t forget about the water and rather simulate the protein. That way, we could reach ten times longer simulations with the same computational effort!
Question: Why do you think that it is important to include explicit solvent in the simulation of a protein?
To check if your assumption is correct, repeat the simulation of protein G, this time without solvent (to observe the effect more clearly, increase the length of the simulation by changing “nsteps” in the file “md.mdp” by e.g. a factor of ten).
Question: What are the main differences to the protein’s structure and dynamics as compared to the solvent simulation?
(Hint: use programs like gmx rms and gmx gyrate to analyze both simulations).
- Let’s go back to the first step in setting up the system – as we already know, building the topology of our protein can be done with the gromacs program pdb2gmx:
gmx pdb2gmx -f 1PGB.pdb -o conf_gromos.pdb
when prompted for the force-field to be used, now choose the GROMOS 43a1 instead of OPLS-AA/L. Use the more and vmd commands as before to compare the result with the previous configuration – what difference do you find?
Question: How is the level of representation correlated with system size (number of atoms)?
Principles of protein structure and basic in biophysics and biochemistry:
- Stryer, Biochemistry
- Voet, Fundamentals of Biochemistry Rev. Ed.
- Cantor and Schimmel, Biophysical Chemistry Part I: The conformation of biological macromoleculesComputer simulations and molecular dynamics:
- M. Karplus and A. McCammon. Molecular Dynamics simulations of
biomolecules Nature structural biology 9: 646-652 (2002).[link]
- D.C. Rapaport. The Art of Molecular Dynamics Simulations – 2nd edn
Cambridge University Press (2004).Advanced reading:
- H. Scheraga, M. Khalili and A. Liwo. Protein-Folding Dynamics:
Overview of Molecular Simulation Techniques Annual Review of Physical Chemistry 58: 57-83 (2007).[link]
- K Henzler-Wildman and D Kern. Dynamic personalities of proteins Nature 450: 964-972 (2007).
- K A Sharp and B Honig. Electrostatic Interactions in Macromolecules: Theory and Applications, Annual Review of Biophysics and Biophysical Chemistry 19: 301-332 (1990).
- F M Richards. Areas, Volumes, Packing, and Protein Structure Annual Review of Biophysics and Bioengineering 6: 151-176 (1977).
- K A Dill, S B Ozkan, M Scott Shell and T R Weikl. The Protein Folding Problem Annual Review of Biophysics 37: 289-316 (2008).
Go back to Contents
For questions or feedback please contact Bert de Groot / firstname.lastname@example.org