Re: [AMBER] Single long pH-REMD simulation vs. several short pH-REMD simulations

From: Cruzeiro,Vinicius Wilian D <>
Date: Mon, 13 Nov 2017 14:51:07 +0000

Hello Eric,

I am not sure what you meant by "convergence issues" on pKa predictions. It is known that implicit solvent CpHMD simulations may yield poor pKa predictions due to an unrealistic conformation sampling that could be done using GB [J. M. Swails, D. M. York, A. E. Roitberg, J. Chem. Theory Comput., 2014, 10, 1341–1352.], however on explicit solvent simulations, unless you have a significant conformational change that happens in a long time scale, once you reach convergence your pKa prediction should not change. You may check that by doing chunk or cumulative analyses on your data.

Based on my experience, I doubt you would get any significant convergence improvement by splitting your pH-REMD simulation into two half-long independent simulations.

Vinícius Wilian D Cruzeiro

PhD Candidate
Department of Chemistry, Physical Chemistry Division
University of Florida, United States

Voice: +1(352)846-1633

From: Eric Lang <>
Sent: Monday, November 13, 2017 6:46:01 AM
Subject: [AMBER] Single long pH-REMD simulation vs. several short pH-REMD simulations

Dear CpHMD experts,

I have a question regarding explicit water pH-REMD simulations with a large number (for example between 40 and 100) of titrable residues.

It might sounds like a trivial question to you but I cannot seem to make up my mind about it...

Based on your experience or tests you have run it is better to run a single long pH-REMD simulation or several shorter ones or are both approaches equivalent?

I guess ideally you want to run several long pH-REMD simulations, each having converged, so you can sample well both the protonation space *and* the conformational space, but because of the computational costs associated with pH-REMD, especially for a titration done over several pH units, it is usually not possible to do so. In such kind of situation is it better to run for example 1 x 500 ns of pH-REMD, 2 x 250 ns, 4 x 125 ns of pH-REMD?

The answer would be obvious in the case of classical MD, but in the case of pH-REMD you want to reach convergence for the titration of ionisable residues which might take long for 'problematic residues' i.e residues that have a poor fit to the Hill equation and/or that interact strongly with one another and thus have interdependent protonation states. So if you run 2 short pH-REMD simulations, you might not reach convergence for both of those simulations, but if you combine them are you obtaining similar pKa estimates than you would for a twice as long single pH-REMD simulation (that has a much better convergence than two separate runs) ?

Multiple pH-REMD simulations will improve the sampling of your conformational space but perhaps it ultimately worsen your pKa estimates because of convergence issues and at the end of the day, it might be better with limited resources to run a single long pH-REMD simulation.

Any insights on the matter would be much appreciated.

Many thanks,


Eric Lang
BrisSynBio Postdoctoral Research Associate Modelling
Centre for Computational Chemistry
School of Chemistry - University of Bristol
Bristol BS8 1TS - United Kingdom
AMBER mailing list
AMBER mailing list
Received on Mon Nov 13 2017 - 07:00:02 PST
Custom Search