• View
• Edit
• History
• Print
• Backlinks

EnergyMinimization

Because of the somewhat low resolution of cryo-EM structures, it is possible for the structure generated from tLEaP to have atoms that have very close contacts. This results in a strong repulsive force, which causes the atoms to rapidly accelerate away from each other once the simulation begins. If the issue is severe, the simulation can crash because the energy calculations result in numbers so high that they cannot be stored in memory. To overcome this issue, an energy minimization protocol is followed.

This begins with all atoms highly restrained so that the highest energy/most problematic atoms adjust their positions first but the rest of the atoms in the system do not move much. This restraint is gradually reduced, changing the restraint reference to the previous structure, over the course of 11 rounds. A combination of steepest descent (a lower energy solution is always chosen) and conjugate gradient (higher energy solutions can be chosen with a certain probability to avoid getting stuck in local energy minima) methods were used.

This protocol was developed by monitoring the total energy of the 5JUP subsystem and then applied to all future energy minimizations. A given round was considered sufficient when the slope of the total energy as a function of the number of steps asymptotically approached 0 (the energy stopped decreasing). The number of rounds was considered sufficient when additional rounds produced little change in overall energy from the previous round. The restraint weight for each round was selected so that each successive round had the same number of steps or fewer.

Energy Minimization Protocol

Energy Minimization sample input

 # 20000 steps of minimization with explicit solvent and ions and 100.0 kcal/mol-A^2 restraints on protein.   
 &cntrl
   maxcyc=20000,	! number of cycles of minimization
   imin=1, 		! energy minimization on
   ntmin=1, 		! switch from steepest descent to conjugate gradient
   ncyc=2500,   	! switch method after 2500 cycles
   cut=9.0, 		! non-bonded cutoff distance
   igb=0, 		! solvent model
   ntb=1, 		! constant volume periodic boundaries
   ntpr=10, 		! report output every 10 steps
   ntr=1,		! restraint on
   restraint_wt=100.0, 
   restraintmask=':1-494',
 &end

Energy Minimization output

  x_emin#.out -- A summary of the system's energy printed incrementally
              -- This file will include information about the run timing at the bottom if the job ran to completion
              -- If the bottom of the file is energy information, the job is either still running or has crashed
  x_emin#.rst -- Single frame containing the atomic coordinates at the end of the minimization round
              -- x_emin1.rst is the input structure for the second round of energy minimization, and so on
              -- x_emin11.rst is used as the starting structure for heating

Energy minimization is run on CPUs. The scripts are already set up for this, so no need to change anything, but this may be helpful information for planning jobs and/or troubleshooting.

Make Energy Minimization Script:
   /home66/kscopino/AMBER18/BIN/make_emin_script.py
   /home66/kscopino/AMBER22/BIN/make_emin_script.py
Tutorial (on Google Drive):
   /ribosome/Molecular Dynamics/Tutorials and Starting Structures/Part2_energy_minimization.mp4

Prev Next