Home >
Community >
How can I run a MD simulation using GROMACS with reverse...
Upvote
24
Downvote
+ Running
+ Molecular dynamics simulation
+ Transition
Posted by
Kasturi Khatun
How can I run a MD simulation using GROMACS with reverse...
Time reversal is not absurd and won’t change the forces (they are proportional to acceleration according to Newton’s second law, not velocities as some ancient philosophers thought) , as a pendulum clearly demonstrates, or qE where E depends on distance. For forces that depend on velocity like qBxv , the vector field B will switch to the opposite direction as it is generated by reversed currents. Stochastic components are different. I would like to be able to change the sign of velocities in the binary trp files, someone knows how to this ? Then make a simulation back and forth and check the numerical accuracy of Gromacs single and double precision as a function of time as the machine introduces stochasticity due to the numerical representation. To have an idea of the accuracy of a 1M step simulation make 500000 steps back and forth or 50 times 10000 steps etc. The displacements must characterize also the accuracy of the 1M step simulation‘s accuracy . I am really curious!
Time reversal is not absurd and won’t change the forces (they are proportional to acceleration according to Newton’s second law, not velocities as some ancient philosophers thought) , as a pendulum clearly demonstrates, or qE where E depends on distance. For forces that depend on velocity like qBxv , the vector field B will switch to the opposite direction as it is generated by reversed currents. Stochastic components are different. I would like to be able to change the sign of velocities in the binary trp files, someone knows how to this ? Then make a simulation back and forth and check the numerical accuracy of Gromacs single and double precision as a function of time as the machine introduces stochasticity due to the numerical representation. To have an idea of the accuracy of a 1M step simulation make 500000 steps back and forth or 50 times 10000 steps etc. The displacements must characterize also the accuracy of the 1M step simulation‘s accuracy . I am really curious!
If you're truly on the top of the free energy barrier from your transition, then running many simulations from this state should give you 50% of forward and 50% of backward transitions. This approach is called committor analysis
If you're truly on the top of the free energy barrier from your transition, then running many simulations from this state should give you 50% of forward and 50% of backward transitions. This approach is called committor analysis
About changing the velocities, you could act on the final configuration and write a script in any programming language to just change the sign of all velocities. I know that this has been done to add a δv to all particles in one specific direction, sorta of giving an instant impulse.
Do you want to change the velocity of all the particles of your system or of some component in particular? In that case you can just use, as before, a script targeting only the component you want to change.
However, due to numerical approximations I think that you cannot achieve perfect reversibility in MD simulations. This is particularly true if you use stochastic couplings, therefore if you are interested in the perfect reversal of your MD trajectory I would say that it makes more sense to just reverse the one you obtained to arrive at your final configuration. Reversing velocities would not take you back along the same phase space trajectory, that's for sure.
About changing the velocities, you could act on the final configuration and write a script in any programming language to just change the sign of all velocities. I know that this has been done to add a δv to all particles in one specific direction, sorta of giving an instant impulse.
Do you want to change the velocity of all the particles of your system or of some component in particular? In that case you can just use, as before, a script targeting only the component you want to change.
However, due to numerical approximations I think that you cannot achieve perfect reversibility in MD simulations. This is particularly true if you use stochastic couplings, therefore if you are interested in the perfect reversal of your MD trajectory I would say that it makes more sense to just reverse the one you obtained to arrive at your final configuration. Reversing velocities would not take you back along the same phase space trajectory, that's for sure.
'Reverse velocities' means reverse forces, and reverse forces means reverse definition of attraction (becomes repulsion) and repulsion (becomes attraction), which is absurd. 'Negative timestep' is a contradiction in terms.
'Reverse velocities' means reverse forces, and reverse forces means reverse definition of attraction (becomes repulsion) and repulsion (becomes attraction), which is absurd. 'Negative timestep' is a contradiction in terms.
Time reversal is not absurd and won’t change the forces (they are proportional to acceleration according to Newton’s second law, not velocities as some ancient philosophers thought) , as a pendulum clearly demonstrates, or qE where E depends on distance. For forces that depend on velocity like qBxv , the vector field B will switch to the opposite direction as it is generated by reversed currents.
Stochastic components are different.
I would like to be able to change the sign of velocities in the binary trp files, someone knows how to this ?
Then make a simulation back and forth and check the numerical accuracy of Gromacs single and double precision as a function of time as the machine introduces stochasticity due to the numerical representation.
To have an idea of the accuracy of a 1M step simulation make 500000 steps back and forth or 50 times 10000 steps etc.
The displacements must characterize also the accuracy of the 1M step simulation‘s accuracy . I am really curious!
Time reversal is not absurd and won’t change the forces (they are proportional to acceleration according to Newton’s second law, not velocities as some ancient philosophers thought) , as a pendulum clearly demonstrates, or qE where E depends on distance. For forces that depend on velocity like qBxv , the vector field B will switch to the opposite direction as it is generated by reversed currents.
Stochastic components are different.
I would like to be able to change the sign of velocities in the binary trp files, someone knows how to this ?
Then make a simulation back and forth and check the numerical accuracy of Gromacs single and double precision as a function of time as the machine introduces stochasticity due to the numerical representation.
To have an idea of the accuracy of a 1M step simulation make 500000 steps back and forth or 50 times 10000 steps etc.
The displacements must characterize also the accuracy of the 1M step simulation‘s accuracy . I am really curious!
More
VOTE
Hi Ankita,
If you're truly on the top of the free energy barrier from your transition, then running many simulations from this state should give you 50% of forward and 50% of backward transitions. This approach is called committor analysis
Hi Ankita,
If you're truly on the top of the free energy barrier from your transition, then running many simulations from this state should give you 50% of forward and 50% of backward transitions. This approach is called committor analysis
More
VOTE
Dear Ankita Katiyar
About changing the velocities, you could act on the final configuration and write a script in any programming language to just change the sign of all velocities. I know that this has been done to add a δv to all particles in one specific direction, sorta of giving an instant impulse.
Do you want to change the velocity of all the particles of your system or of some component in particular? In that case you can just use, as before, a script targeting only the component you want to change.
However, due to numerical approximations I think that you cannot achieve perfect reversibility in MD simulations. This is particularly true if you use stochastic couplings, therefore if you are interested in the perfect reversal of your MD trajectory I would say that it makes more sense to just reverse the one you obtained to arrive at your final configuration. Reversing velocities would not take you back along the same phase space trajectory, that's for sure.
Dear Ankita Katiyar
About changing the velocities, you could act on the final configuration and write a script in any programming language to just change the sign of all velocities. I know that this has been done to add a δv to all particles in one specific direction, sorta of giving an instant impulse.
Do you want to change the velocity of all the particles of your system or of some component in particular? In that case you can just use, as before, a script targeting only the component you want to change.
However, due to numerical approximations I think that you cannot achieve perfect reversibility in MD simulations. This is particularly true if you use stochastic couplings, therefore if you are interested in the perfect reversal of your MD trajectory I would say that it makes more sense to just reverse the one you obtained to arrive at your final configuration. Reversing velocities would not take you back along the same phase space trajectory, that's for sure.
More
VOTE
'Reverse velocities' means reverse forces, and reverse forces means reverse definition of attraction (becomes repulsion) and repulsion (becomes attraction), which is absurd.
'Negative timestep' is a contradiction in terms.
'Reverse velocities' means reverse forces, and reverse forces means reverse definition of attraction (becomes repulsion) and repulsion (becomes attraction), which is absurd.
'Negative timestep' is a contradiction in terms.
More
VOTE