Exercise 5: Proton transfer in an Enzyme
This examples shows how to setup/run a model system for a 'real-life' simulation.
The final part of this tutorial shall demonstrate how to set up a model system for a 'real-life' simulation. You can find some background information about this model in this extract from the Bachelor Thesis of Rachel Glaves.
Refer to the PDF document for more details.
Preparing a Model from a Large System
Requirements: Memory: 100 MB, CPU time: 10 min.
We want to look at the active sites in an acetylcholineesterase enzyme. The full enzyme, especially when including the solvating water molecules, is by far too large to be treated with Car-Parrinello MD. Thus the fist step is to extract a model system from a fully solvated and equilibrated peptide. The main idea is to extract a number of residues, only keep the amino acid sidechains and the alpha-carbons which are turned in to methy groups and fixed in space. To do this manually can become quite cumbersome, but the scripting capabilities of VMD will help us to achieve our goal. The file AChE.gro contains the solvated and equilibrated enzyme. The VMD script get-triade.vmd will now extract the coordinates of the three residues that comprise the triade and write them to a PDB file (reftrajectory.pdb). To run this script you can either start first start VMD and then load the script via source get-triade.vmd from the VMD terminal prompt or have VMD execute the script on loading via vmd -e get-triade.vmd . Since we don't need the graphical user interface for pure script processing, you can also use the -dispdev none option to VMD to disable the GUI and the OpenGL window. The resulting PDB file will also contain box size information that will add a 2 Angstrom safety margin to avoid overlapping atoms via the periodic boundaries. Change the script to extract only the HIS440 residue and run a geometry optimization for the positions of the hydrogens only with CPMD (with ultra-soft pseudopotentials and using FIX ATOMS).
Equilibration with a Blocked Reaction Path
Requirements: Memory: 400 MB, CPU time: 20+15 min.
For the following steps we use the minimal triade subsystem with the acetlycholine added (and bound) and including 4 water molecules. Since we are looking at an intermediate state of the enzymatic reaction, yet we have to equilibrate the model system, as we have removed most of the atoms surrounding the active site. So we have to find a way to inhibit the reaction. In this case adding a proton and turning the ester group into an semi-acetal group does the trick nicely. Use 1-prot-triade-wfnopt.inp to calculate the optimized wavefunction and then start an MD run in the usual way (2-prot-triade-equilib.inp). In this case, we run only a few steps and skip the further equilibration (3-prot-triade-cont.inp). In fact, the start configuration was taken from a fully equilibrated configuration. An xyz-movie file (TRAJEC.xyz) of such an equilibration run is available.
Modelling Part of the Reaction Path
Requirements: Memory: 450 MB, CPU time: 45+60 min + 17h Production.
Now remove the inhibiting proton from the previous wavefunction optimization input, don't forget to adjust the rest of the input (CHARGE!) and start the 'real thing'. Equilibrate with (TEMPCONTROL IONS 300.0 20.0) for 60 steps and then switch to Nose-Hoover chains with the following parameters:
NOSE IONS MASSIVE 300.0 2500.0 NOSE ELECTRONS 0.007 15000.0 NOSE PARAMETERS 3 3 3 6.0D0 15 4
Now we need run the simulation at least until the next day to see a reaction to happen. Depending on the circumstances it will be either the forward or backward reaction. If you use the MAXCPU keyword you can set MAXSTEP as high as you want, after the indicated period the simulation will stop and write a restart. For instance with a value of 60000 for MAXCPU the job should stop after about 17 hours. This can also be extremely useful for running in a batch environment (especially in combinations with the RESTFILE and STORE keywords).
Calculating Electron Structure Properties and Visualizations
Requirements: Memory: 1200 MB, CPU time: 3-5 h/WFopt.
Since we were using ultra-soft pseudopotentials, we have to recalculate the electron structure with norm-conserving pseudopotentials (at least for most of the properties). Since this will need a lot of memory and CPU resources, already pre-calculated files are provided in the references output directory.