(Note: These tutorials are meant to provide illustrative examples of how to use the AMBER software suite to carry out simulations that can be run on a simple workstation in a reasonable period of time. They do not necessarily provide the optimal choice of parameters or methods for the particular application area.)
Copyright Ross Walker 2004

TUTORIAL 6 - AMBER 8 VERSION - SECTION 3

A Coupled Potential QMMM Simulation

By Ross Walker

PLEASE NOTE THIS TUTORIAL IS FOR AMBER 8 ONLY.
AN AMBER 9 QM/MM TUTORIAL IS AVAILABLE HERE

3) QM/MM MD

Now, let's try repeating the simulation using a coupled QMMM potential. In this run we will use sander.QMMM instead of sander. We will model the NMA molecule using the semiempirical PM3 Hamiltonian while all the water we will model classically. In this situation we have no bonds crossing the QMMM boundary and so do not need to worry about where sander.QMMM will place hydrogen link atoms. There will be none.

For QMMMMD the input files look very similar. There are a few notable differences, however, related to the specification of the atoms to be modelled quantum mechanically. Here is the minimisation input file we will use:

The extra variables in this file are:

ifqt = 1 - This is the flag that tells sander.QMMM that we want a QMMM run.
nqt = 12 - The number of quantum atoms - in this case the whole NMA molecule = 12.
idc = 0   - We want to the standard SCF procedure. NOT divide and conquor.

We also have a list of numbers after the / that ends the &cntrl namelist. These numbers specify the atoms that we want included in the QM part of the calculation. There should be nqt of these numbers.

We also need a second input file that tells the QM code what options we wish to use. This must be placed in the directory in which we will run the simulation and it must be named divcon.in. For the moment a divcon.in file must be present for all sander.QMMM runs. In AMBER 9 this file is no longer required and instead all options are specified in the mdin file.

An explanation of the options in this file are as follows:

CARTESIAN - Tells divcon that we will be providing it with cartesian coordinates.
PM3 - We will be using the semi-empirical PM3 Hamiltonian.
CHARGE=0 - Tells divcon that the charge of our QM system is neutral.
STANDARD - We will be doing a standard (non divide and conquor) SCF.
DIRECT - Do a direct SCF calculation. Do not store integrals on disk.

As we did in the classical simulation above we will run 1,000fs of simulation. Again our input file is the same apart from the QM specifications. We will use the same divcon.in file for both the minimisation and MD. Here is the MD input file:

Now we can run the QMMM minimisation and MD in the same fashion as we did for the classical simulation (remember, however, that we have to use the sander.QMMM executable and NOT sander here):

$AMBERHOME/exe/sander.QMMM -O -i min_qmmm.in -o min_qmmm.out -p NMA.prmtop -c NMA.inpcrd -r NMA_min_qmmm.crd

$AMBERHOME/exe/sander.QMMM -O -i md_qmmm.in -o md_qmmm.out -p NMA.prmtop -c NMA_min_qmmm.crd -r NMA_md_qmmm.rst -x NMA_md_qmmm.mdcrd

You can track the output in another terminal window with "tail -f md_qmmm.out" if you wish. The first thing you should notice with this simulation is that it takes significantly longer than the classical simulation. The reason for this is that we are doing a complete SCF at every step. The performance should be increased significantly in AMBER 9 when we will be using a number of optimisations to this process to avoid having to do as much work as AMBER 8 does at the moment.

These calculations should take around 7 mins to run. Here are the output files: min_qmmm.out, NMA_min_qmmm.crd, md_qmmm.out, NMA_md_qmmm.mdcrd