AMBER TI Tutorials
Introduction to TIMD
This is a TIMD tutorial based on the tutorial written by Thomas Steinbrecher. But some important changes have been made to suit the current AMBER 10 version according to Miranda's tutorial from Simmerling's lab.
And this is my own version of TIMD of the T4-L99A enzyme, the results are a little different from Thomas' results due to some changes in the parameters. Always remember that the experimental value is the absolute criteria.
In this tutorial, free energy calculations will be used to calculate the relative binding free energy of two simple ligands, benzene and phenol to the T4-lysozyme mutant L99A. Free energies will be computed by using the thermodynamic integration facilities of the sander program. A modified van-der-Waals equation (softcore potentials) are used to ensure smooth free energy curves.
Thermodynamic cycle and Method
TI calculations compute the free energy difference between two states A and B by coupling them via a parameters λ that serves as an additional, nonspatial coordinate. This λ formalism allows the free energy difference between the states to be computed as:
From the pictures above, you can see that Processes A and B represent the binding of two different ligands to a protein, while processes C and D are transformations from one ligand to the other while it is bound to the protein (C) or simply solvated in water (D).
Since Δ GC-Δ GD = Δ GA-Δ GB, TI calculations can be used to compute relative binding free energies, making them useful tools in drug design or lead optimization applications.
Preparation for setup of the T4 L99A System
The two ligands were sketched and parametrized with gaff atom types and resp charges were generated using antechamber on gaussian03 output files. (Please refer to a basic AMBER tutorial on how to use the antechamber tools to parametrize a ligand. The benzene and phenol molecules were saved in two OFF-libraries (benz.lib and phen.lib) for further use. The screenshot shows that C6 was selected as the position bearing the hydroxyl group in phenol.
We are going to use the X-ray structure of T4-L99A from the pdb (after stripping water molecules and unneeded heteroatoms from it: pdb file) as basis to set up our simulation files. Generally, you can strip the waters and other unneeded molecules in another software, but here the modified pdb file is already provided.
We will use two runs of leap to produce four sets of parameter and restart files, containing both ligands in the protein bound and solvated states.
Notice if there are sharp changes in the TIMD process, more sets of parameter and restart files are required(generally 6 sets). For example, in Miranda's tutorial, she has 3 sets of topology/coordinate files for each transformation: two 'original' endpoint files generated directly from crystal/MD structures, one 'fake' endpoint file generated by mutating one 'original' endpoint file. Thus totally 6 sets of files for her tutorial.
tleap to generate sets of parameter and restart files
We will use two runs of leap to produce four set of parameter and restart files, containing both ligands in the protein bound and solvated states. The first leap run (input file) will produce pdb files of the solvated und neutralized benzene complex and of the benzene ligand in water (complex.pdb and ligand.pdb). From these two additional pdb files are made by renaming the BNZ molecule to PHN and deleting H6 (t4_phn.pdb and phn.pdb).
The script for first tleap run:
source leaprc.ff03 source leaprc.gaff loadoff benz.lib
complex = loadpdb 181L_mod.pdb solvatebox complex TIP3PBOX 12 addions complex Cl- 0 savepdb complex complex.pdb
ligand = copy BNZ solvatebox ligand TIP3PBOX 12 savepdb ligand ligand.pdb
These four pdb files are then used in a second leap run (input file) to generate the *.prm and *.rst files. This yields 4 parameter and 4 rst files.
The script for second tleap run:
source leaprc.ff03 source leaprc.gaff loadoff benz.lib loadoff phen.lib
complex = loadpdb complex.pdb setbox complex vdw savepdb complex t4_bnz_leap.pdb saveamberparm complex t4_bnz.prm t4_bnz.rst
ligand = loadpdb ligand.pdb setbox ligand vdw savepdb ligand bnz_leap.pdb saveamberparm ligand bnz.prm bnz.rst
complex = loadpdb t4_phn.pdb setbox complex vdw savepdb complex t4_phn_leap.pdb saveamberparm complex t4_phn.prm t4_phn.rst
ligand = loadpdb phn.pdb setbox ligand vdw savepdb ligand phn_leap.pdb saveamberparm ligand phn.prm phn.rst
Start Structures Results
After the following steps, you should already have 4 sets of starting structures for the system. Now we should focus on the scripts used to run the thermodynamics integration.
There are totally 6 directories for you to make a 3-step change in 2 different environments. Moreover, in each directory, you should do 3 steps for a complete simulation: minimization, equilibration, production.
Setup and run MD
I would create 6 directories to run the Thermodynamics Integration Molecular Dynamics separately.
3 steps: removing the charges from benzene, VDW change from benzene to phenol, adding the charges to phenol. 2 environments: unbound water environment and bound with T4 L99A environment.
For the second step which is VDW change from benzene to phenol, some important changes have been made. In AMBER 10, a new kind of method is used: softcore potentials algorithm. Using this algorithm removes the requirement to prepare "dummy" atoms and allows two parameter/top files to have different atoms.
Several things to notice when run the molecular dynamics:
- you can use different numbers of windows. But when you use the new windows, the original windows' simulation data could also be used.
- running the minimization separatesly for the VDW change step. Because in this step, only 2 processors could be used to run the minimizations. You can use more processors for the equilibrium and production steps.
- Make sure to always use Langevin when you are applying soft core potential, for this is a requirement in AMBER manual.
- ntf should equal 1 for all softcore part. Otherwise, there would be errors or crashes in the SANDER module.
the bash scripts for running the simulations are listed below:
After running the simulations on the seawulf cluster, you still have plenty of work to do. The analyze of the data is not so straight and you need to write linux scripts using python or perl to extract the data.
I write a python extraction scripts myself to extract the data DVDL and rms data like the following format: