MMTSB Tool Set - Analysis of replica exchange simulations

Objective and Overview

This tutorial will illustrate how to use the MMTSB Tool Set to analyze replica exchange simulations, including weighted-histogram analysis to obtain free energy maps.

In order to generate the input data for this tutorial it is necessary to run the replica exchange tutorial (WALP peptide) first. It is assumed that the output from that tutorial is available and that you are entering the commands in this tutorial from the main replica exchange output directory.

1. Information about replica exchange simulation

The tool can provide various output related to the progress of the replica exchange simulation (rather than the sampled conformations which we will look at later).

For example, it is possible to examine how the temperature varies for a given replica with: -byclient aa1
"Client" means replica here, and for all-atom replica exchange simulations, the replicas are named aa1 through aaN. The output from the above command contains the temperature in the fifth column, and the internal energy in the sixth column. The RMSD with respect to a reference structure may be available in the seventh column if a reference structure was provided during the replica exchange run.

Extract the variation of temperature for each of the four replicas and plot the results to examine whether each of the replicas samples the entire range of temperatures (as it should).

Another way to query the replica exchange output is to look at which clients are at the lowest temperature at each cycle: -bytemp 300
The output from this command is very similar as in the previous case. How does the RMSD from the reference evolve at different temperatures?

Often, it is most interesting to find out which structures are sampled most at the lowest temperature as these conformations have the lowest free energies. The following command examines the last 100 cycles and determines at which temperature different replicas spend most of their time: -ranklast 100
The output from this command shows for each replica the average rank in terms of temperature (with 1 corresponding to the lowest temperature) as well as the percentage of time spent at the lowest temperature. The conformations from the replica with the lowest average rank and largest percentage spent at the lowest temperature are most favorable at the lowest temperature.

2. Extraction of single conformations

So far we have not looked at any actual structures. The sampled conformations (saved at the end of each replica exchange cycle) are stored in a subdirectory for each replica (aa1 to aaN). If the archive parameter was specified during the replica exchange run, all of the conformations are stored in a single file (prod.coor.archive). They can be accessed with An example of how to use this tool is given in the following: -inx 100 -xtract -pdb aa1/final.pdb aa1/prod.coor.archive This command extracts the conformation after the 100ths cycle from replica aa1. The extracted structure is written out in PDB format to a file set100.pdb.

Use this command in combination with to extract the structure from the most favorable replica at the lowest temperature after 200 cycles.

3. Analysis of multiple conformations

The tool can also be used to analyze a number of conformations from a given replica. For example, the following command calculates the radius of gyration for the conformations that are sampled by the first replica during the first 200 cycles: -inx 1:200 -apply \
                   -pdb aa1/final.pdb aa1/prod.coor.archive
Because operates on single archive files, it cannot easily analyze properties as a function of temperature instead of replica. A different, more general tool,, is available for that purpose. It is used as follows: -inx 1:200 -bytemp 300 -apply
In this example, the radius of gyration is calculated for the conformations - from different replicas - that are found at T=300K during the first 200 cycles.

As in the previous command the radius of gyration calculation is performed with the external command. In fact, any program or script could be used instead. This provides a lot of flexibility but it is not particularly efficient for the analysis of large replica exchange simulations. Another option is to provide a perl function analyze that takes a Molecule object as an argument and returns a value.

In the radius of gyration calculation example such a function would look simply like this (in analogy to the script):
    sub analyze {
     my $mol=shift;
     return &Analyze::radiusOfGyration($mol,1);
Enter this piece of perl code into a file named rgyr.function and rerun the replica exchange analysis tool: -inx 1:200 -bytemp 300 -function rgyr.function  
This command should give essentially the same output at a much faster speed because the time for starting an external program for each conformation is saved.

4. 2D WHAM free energy analysis

A powerful method for analyzing the results from replica exchange simulations is weighted-histogram analysis that is most convenient and informative when based on two reaction coordinates that are appropriate for distinguishing the sampled conformations.

In order to prepare the input for the WHAM analysis we need the values of the two reaction coordinates for each temperature at each replica exchange cycle that is being used for analysis. The tool generates such data if a specific temperature is not given: -inx 50:500 -function rgyr.function >
In this case we will skip the first 50 cycles that are considered equilibration. Use in combination with to also generate the percentage of helicity as a second reaction coordinate. While there are different ways to complete this task it is best to write another function based on that calculates the helicity in the following command: -inx 50:500 -function helicity.function >
Once we have the data for the reaction coordinates we can run the WHAM analysis with the following command: -wham \
                   -whamtemp 300:400
The arguments to the -wham option describes the reaction coordinates, including the desired range and binning for the histogram generation. The temperatures at which the PMF surfaces are generated are given with -whamtemp.

As a result of this command a number of output files are generated in the current directory. The most interesting are the one and two-dimensional PMF profiles in the files T=xxx_a_vs_rgyr.out (1D) and T=xxx_a_vs_rgyr_vs_helicity.out (2D).

The 2D profile can be plotted with gnuplot with the following commands:
    set cntrparam levels 0,1,2,3,5,10
    set dgrid3d 10,10,2
    set contour base
    splot "T=300_a_vs_rgyr_vs_helicity.out" us 1:2:3 w lines
Compare the surfaces between 300 and 400K. What do you learn from this?

Written by M. Feig