Is it possible to move all atoms simultaneously in each timestep?

Hi everyone,

As you know, in MD simulation of a biological system, typically, one atom is moved in each timestep, followed by recalculating forced, then another atom is moved and so forth. Does GROMACS, or any other MD simulation tool such as NAMD, AMBER, etc., allows us to run the simulation such that all atoms move simultaneously in every single timestep of the simulation before recalculating forces and repeating the process? There was a study that showed an more accurate results when doing so.

Thank you in advance!

–Ra’ed

I’m not sure I understand what you mean. In MD simulations all forces on atoms are calculated every time step in order to update their velocities. All atoms are moved simultaneously according to their velocities. Did you mean something else?

If you mean Reversible reference system propagation algorithm (r-RESPA), It updates the fast interactions more often and the slower ones less often, offer efficiency benefits without sacrificing much accuracy. It also maintains reversibility, meaning the algorithm allows for backward integration to retrace the simulation steps accurately.

GROMACS doesn’t directly implement this algorithm in its core functionalities. Instead, it relies on other integration schemes that optimize the simulation process while maintaining accuracy. for example, you can use multiple time-stepping algorithms (e.g., velocity Verlet with different time steps for bonded and non-bonded interactions) that help speed up simulations without compromising accuracy significantly.

Thank you so much, Farzane, for the elaboration on that, but this is not what I meant.

Thank you so much, Magnus, I was under the impression that one atom is moved, followed by calculating only the forces impacted by that move, then another atom is moved, and so forth. There was an article stating that moving all atoms is rather closer to what happens in real life and provided better results. I wonder if they were referring to other MD simulation tools. Let me see if I can find the article and get back to you.

No, that would be very inefficient with many atoms and the concept of time during the simulation would be very strange.

The forces on all atoms (from all atoms, including short-range and long-range interactions) are calculated, all velocities are updated and all atoms are moved every timestep.

Thank you for confirming, Magnus! I appreciate it.