Parallelizable Gaussian Accelerated Molecular Dynamics (ParGaMD)

ParGaMD is a hybrid method that leverages the accelerated sampling of Gaussian Accelerated Molecular Dynamics (GaMD) and marries it with the excellent parallelization of the Weighted Ensemble (WE) framework, resulting in a powerful method that can speed up protein simulations.

Please follow the given steps in order to successfully perform a ParGaMD simulation:

1. Perform a conventional Gaussian Accelarated Molecular Dynamics (cGaMD) simulation in the cMD folder by running sbatch run_cmd.sh to get the GaMD paramters (gamd-restart.dat)
2. The run_WE.sh in the main directory takes care of copying gamd-restart.dat and bstate.rst into the required directories in the WE framework.
3. Make sure that pcoord_len = nstlim/ntpr + 1 in west.cfg (nstlim, ntpr are in common_files/md.in).
4. Run the command squeue -u username to copy the job ID of the cGaMD simulation. This job ID will be used in the next step.
5. Run the ParGaMD simulation as a dependancy job by running the command sbatch --dependency=afterok:jobid run_WE.sh . (Please change the NODELOC in env.sh to the directory from where you're running the simulations)
6. After running the simulation, run the postprocessing to get the gamd.log and PC.dat (PC --> progress coordinate) files using run_data.sh (NOTE: Please submit run_data.sh on to the compute node and not on the login node as there may be memory allocation issues).
7. For reweighting to get the free energy surface (FES):
   1. Get weights.dat from gamd.log using the command awk 'NR%1==0' gamd.log | awk '{print ($8+$7)/(0.001987*300)" " $2 " " ($8+$7)}' > weights.dat
   2. Get output.dat from PC1.dat and PC2.dat (If using 2 PCs for a 2D surface) using the command awk 'NR==FNR{a[NR]=$2; next} {print a[FNR], $2}' PC1.dat PC2.dat > output.dat
   3. Run the command ./reweight-2d.sh 50 50 0.1 0.1 output.dat 300, where 50 50 are the cutoffs for both PCs, 0.1 0.1 are the bin spacings for the PCs, and 300 is 300 K.