Rst-wham-2d: Difference between revisions

The command line arguments largely have the same meaning as they do for the one dimensional [ rst-wham ] program.

|<span style="color: blue;">metadata</span><br/><span style="margin-left: 0px;">Each line of the metadata file should be blank, begin with a "#" (marking a comment), or have the following format:

::<code>/path/to/timeseries/file   loc_win_min  spring  [correl time] [temperature]</code>

This first field is the name of one of the time series files (more on this in a moment). The second field, loc_win_min, is the location of the minimum of the biasing potential for this simulation, a floating point number. The third field, spring, is the spring constant for the biasing potential used in this simulation, assuming the biasing potential is of the format

::<math> V= \frac{1}{2}k (x-x_0)^2 </math>

Many simulation packages, including TINKER, AMBER, and CHARMM, do not include the [1/2] when they specify spring constants for their restraint terms. This is a common source of error (I'd love to change my code to match the other packages' behavior, but then experienced users who don't read the manual would get messed up). Also, the units for the spring constant must match those for the time series. So, if your time series is a distance recorded in Ångstroms, the spring constant must be in kcal/mol-Å2. AMBER users should take care when using angular restraints: the specification and output of angles is in degrees, but AMBER's spring constants use kcal/mol-rad^2.

The fourth argument ("correl time") specifies the decorrelation time for your time series, in units of time steps. It is only used when generating fake data sets for Monte Carlo bootstrap error analysis, where it modulates the number of points per fake data set. This argument is optional, and is ignored if you don't do error analysis. If you're doing multiple temperatures but not bootstrapping, set it to any integer value as a placeholder, and it'll be ignored. See section for more discussion about how to do bootstrapping.

Finally, the last (optional) field is the temperature for this simulation. If not supplied, the temperature specified on the command line is used. In the present version of the code, you must either leave the temperature unspecified for all simulations or specify it for all simulations.

If the simulation temperature is specified, there must be a third column of data, containing the system's potential energy at that time point. It should be a floating point value.  

The first line of the output file contains echoes command line. The next line or two contain comments describing the periodicity used and the number of simulation windows present. While the calculation is running, it will print out lines that look like the following:

  #Iteration 10:  0.106019

  #Iteration 20:  0.062269

  #Iteration 30:  0.039890

  #Iteration 40:  0.027003

This specifies the current iteration number, and the average change in the F values for the current iteration. This number is not used for deciding when the calculation has converged; rather, the maximum change, as opposed to the average, is used.

Every 100 iterations, the current version of the PMF is dumped into the output file. These lines look like

 

 

The first column is the value of the reaction coordinate, the second is the value of the PMF, and the third is the unnormalized probability distribution.

  # 0     0.000004

  # 1     -4.166136

  # 2     -3.241052

  # 3     -4.475215

# 5    -7.128731

  #Coor          Free            +/-             Prob            +/-

-178.000000     0.014386       0.000098       0.106389       0.000017

  -174.000000     0.068560       0.000151       0.097128       0.000025

  -170.000000     0.250825        0.000350        0.071496        0.000042

-166.000000     0.523786       0.000294       0.045186       0.000022

The first column is the value of the reaction coordinate, the second is the free energy. The third is the statistical uncertainty of the free energy (which is only meaningful if you performed Monte Carlo bootstrapping). The fourth and fifth columns are the probability and it's associated statistical uncertainty. Again, the latter is only meaningful if bootstrapping is performed. See section for further discussion of error estimation.