What are the accepted values for the exchange probability of REMD simulation?

I tried more than one time and got these values, how can I know whether values are accepted or not for the exchange probability of REMD simulation ?

Is there any help please

the output reports the exchange probability and number of exchange between 2 replica (in your case the probability is between 23-40 % and number of exchange is between 300-600). The exchange probability is linked to how you defined the replicas (see more here Replica exchange — GROMACS 2021.4 documentation). an exchange acceptance probability around 0.2 is usually a good criteria, I guess you can read more on it in literature.
I hope this answer your questions

1 Like

Thanks a lot for replying. excuse me , does it mean I have to change the temperature range in the replica which gives a 0.4 exchange probability?


a change in the temperature range will affect the expected exchange probability as described in the formula in the manual.
Usually one does not want to exchange neither too fast (system do not have time to relax at that temperature) neither too slow (weak sampling). Works in literature suggest to have an exchange probability of around 20%.

In your specific case, it depends on your specific problem (e.i if you are interested in all the temperatures or at a specific temperature). Maybe you can first analyze the results.


1 Like

Thanks a lot for replying . I’m trying to find the accurate confirmation structure for my peptide so I did the REMD simulation , however, I’m following the attached screenshot in this paper , he did analysis for only trajectory 300K so, in this case, i have to concatenate all frames or just deal only for 300K

is there any response please ?

If you have a question about a published paper and how to reproduce it or follow the same method, contact the authors of the paper.

thanks but it tried without any response so I’m asking here if there is someone who can help … appreciating your time d

Normally one only ever analyzes the ground-state replica as the others are artificial, anyway.