If I simulate the two similar systems with similar parameters under different CPU load and nodes/processors. Then, will I get the replicate results ?
I am asking this to be ensure as I am going to simulate four protein ligand complexes using different machine with different hardware. So, while comparing their relative binding energetic, is it all right to compare them or else is it necessary to simulate all under exactly similar hardware capacity/load for parallel runs?
Try and see. If you get the same coordinates by running a 2-step MD simulation under different hardware conditions (you will get the answer in a heartbeat) you will also get the same energies.
MD simulations have a chaotic behavior, so it's expected the results will be different for the same systems, running on different CPU cores.
With experience in replicate studies i would say yes (assuming you are going to start the replicate simulation from scratch and not from the previous run's binary) and also its not just that, even if you replicate on the same hardware with same system (parameters/setup) you will certainly experience difference from one replicate to the other, the reason is very simple which you can find from these below threads.
https://www.researchgate.net/post/Why_does_performing_molecular_dynamics_simulation_on_different_CPUs_lead_to_different_results
http://www.gromacs.org/Documentation/Terminology/Reproducibility
- Badges
- Science topic