Tutorial For Well-Sliced Metadynamics

 

Primary Reference:

See the paper: click here

 

Shalini Awasthi, Venkat Kapil and Nisanth N. Nair  “ Sampling Free Energy Surfaces as Slices by Combining Umbrella Sampling and Metadynamics'' J. Comput. Chem.   37, 1413-1424 (2016).

 

 Algorithm

 

 Here, we briefly explain the algorithm for the well-sliced metadynamics (WS–MTD) approach. This technique requires no special implementation in a molecular dynamics.

 code which can carry out well-tempered metadynamics (WT–MTD) simulation and umbrella sampling (US) runs simultaneously for different collective variables (CVs).

 

The reweighting procedure works as a post processing.

 

1. For the chosen range of values of  CVs, place  restraining umbrella potentials. For every umbrella, carry out a WT–MTD simulation sampling the  coordinates.

From these simulations, obtain the time series of  ,  for some regular intervals of MTD time .

2. Compute   for . 

3. Compute   for .

4. Plot , and based on that choose a time range,  and , for which , and a proper sampling of  has been accomplished.  

5. For the time range, construct the MTD–unbiased distributions.

6. Using WHAM, reweight the umbrella potential as well as combine the  distribution functions to get .

  Note that  can be the input for any standard WHAM programs, but by setting the bias along the  coordinates to zero.

7. Construct the free energy surface .

 

Alanine-dipeptide

 

To study alanine–dipeptide, we use molecular dynamics using a Langevin thermostat at 300 K, with a time step 1.0 fs, using AMBER force–field. We employed the PLUMED–AMBER interface for carrying out WS–MTD runs (input files: plumed_input_file, amber_input_file, topology_file, coordinate_file).

In WS–MTD simulations,  coordinate is sampled using WT–MTD and  was sampled using US. The choice of coordinates here is arbitrary. However, the US coordinate is the one along which you are expecting a broad and/or unbound free energy basin. 

 

Here, Gaussian potentials are updated at every 500 fs, and  kcal mol,  radians and  K are taken. Umbrella potentials are placed from  to  at an interval of 0.2 radians with  kcal mol rad.

Details of the input files required for simulation can be found elsewhere.

 

Reweighting

Compile metadynamics unbiasing code.

$ gfortran mtd.f -o mtd.x

 

Prepare the input files (COLVAR, HILLS, run.sh).

If one is using PLUMED for running simulations output files can be directly used.

(Remove commented lines from COLVAR and HILLS file.)

 

COLVAR file should have following information

 

HILLS file should have the following information:

 

 

run.sh has following information:

                                                                                                                                                             

Run command run.sh for processing files from different umbrella windows.

$ sh run.sh

Output file: Probability file (Pu.dat). Rename it as PROB_i, where “i” is the umbrella index.

These PROB files contain values for CV1, value for CV2, and probability of (CV1,CV2).

PROB files should be plotted to see the overlap between different umbrellas. If one finds insufficient overlap, then

more umbrellas can be simulated for intermediate values.

 

Collect all the PROB_i files in a folder PROB.

These PROB_i are the input for WHAM analysis.

 

Compile the WHAM code

$ mpif90 wham.F90 -o wham.x

 

Prepare WHAM input files (WHAMINPUT, input).

input file contain following information:

Temperature of physical system should be provided in Kelvin. Here, grid information should be same as used during MTD unbiasing.

 

WHAMINPUT file contains following information:

 

Run the executable (wham.x).

$ mpirun –np 8 wham.x  (Running on 8 processors)

or

$ ./wham.x

 

Input file: PROB_i, WHAMINPUT, input

Output file: free_energy, which contains the free energies.

Free energies can be plotted using gnuplot using fes.gnp .