Re: [AMBER] Heating the soft core atoms in TI simulations

From: He, Amy via AMBER <>
Date: Wed, 25 Jan 2023 23:43:46 +0000

Dear Dr. York and Dr. Steinbrecher,

Thanks so much to you both for your kind replies. That is really helpful.

I got the temperatures from the MD output (printed along with the energies). I now understand that the temperatures are theoretical temperatures representing the kinetic energies of the atoms. Because I have only 3 soft core atoms, the fluctuations are large but I don’t think they are constantly or unreasonably lower or higher than the target temperature.

I have just two more questions:

1. Regarding the barostat option in TI simulations

I read from previous discussions on our mailing list that, barostat = 2 (using the Monte Carhlo barostat) is recommended for TI simulations instead of the default Berendson barostat.

I am using Amber20 and sander.MPI to run my current calculations. However, when I tried barostat = 2, it says “TI is not compatible with the MC Barostat”.

As the only option in sander, is barostat = 1 still ok for TI simulations? Can you please point me to a discussion or literature on why Monte Carlo barostat is preferred?

2. Regarding the ACES protocol, or setting up a “lamda-REMD” calculation in Amber

For a transition with major conformational change, it would be super helpful if the alchemical free energy calculation can be complemented by REMD. Could you please point me to a tutorial on how to use the ACES protocol, or similarly, setting up a “lamda-REMD” in Amber? I have been thinking about doing this for a long time.

Thanks again for your advice. I really appreciate your help and kind suggestions.


From: Darrin York via AMBER <>
Date: Wednesday, January 25, 2023 at 11:02 AM
To: <>
Subject: Re: [AMBER] Heating the soft core atoms in TI simulations
Hi Amy,

Thomas is a real expert at this stuff and is spot on - I'll just to add
a little bit in the context of some new features that have been recently
implemented in Amber that might be useful.

As Thomas mentioned - sometimes strategically including more atoms into
the softcore region can serve to enhance sampling, although often at the
expense of slower convergence (or the requirement of having more
windows). This idea is this: suppose that as a consequence of your
alchemical transformation your side chain adopts a quite different
conformation that involves torsion angles throughout the side chain
(involving more atoms than one might minimally use as your softcore
region). If you choose your softcore atoms to be "minimal" - then
during your alchemical transformation, you must also sample the physical
transition from one conformational state into the other and suffer
whatever barriers are encountered in that change.

The ACES enhanced sampling method that was recently introduced
(;!!KGKeukY!2HVRmvKMxeBHFS3EoaMhud9k4msPmq5XFc_Innmcn4Kgm4wtK1Fc-C0hAHNQiFWorBed91caSLCsCrW2o4q-56QSrck$ ) leverages the dual
topology and Hamiltonian replica-exchange framework within Amber, along
with some new forms for the softcore potential
(;!!KGKeukY!2HVRmvKMxeBHFS3EoaMhud9k4msPmq5XFc_Innmcn4Kgm4wtK1Fc-C0hAHNQiFWorBed91caSLCsCrW2o4q-kXIjJSY$ ) to help
overcome this type of problem. In essence, instead of having to sample
the physical conformational transition, ACES uses alchemical paths to
make the one conformational state smoothly disappear while the other
state smoothly appears - effectively tunneling through the physical
conformational barrier.

I am not sure if this situation applies to your case, but I thought I
would mention it.

best regards,


On 1/25/2023 2:28 AM, Steinbrecher, Thomas via AMBER wrote:
> Hi Amy,
> what does 'not a stable temperature' mean here? It is normal for the
> ill-defined temperature of a few atoms to fluctuate wildly, but the (very)
> long run average should still be around your target temperature. So if the
> SC atoms are at 400 +- 200 K during the simulation, I would not worry much,
> if they are at 1K or 10000K, something would indeed be wrong.
> There is no theoretical limit for the softcore/transformation region of a
> free energy calculation, but your convergence slows drastically once you go
> above a handful of atoms. Including a whole sidechain may make sense, but
> only if the alchemical change in the side chain will also impact its
> preferred conformation. Otherwise, I'd recommend keeping SC regions as
> small as you can (Though making it bigger will help make the temperature
> look more normal, if nothing else).
> Kind Regards,
> Thomas
> On Wed, Jan 25, 2023 at 2:01 AM He, Amy via AMBER <> wrote:
>> Dear Amber Experts,
>> I have a question about Amber TI simulations with soft core potential.
>> I want to do a single step morphing and calculate the energy difference of
>> changing the charge states of two residues in a protein.
>> My system is solvated with explicit waters. There are three soft core
>> atoms: The appearing/disappearing H, and two nearby atoms that have
>> different types in different charge states. SHAKE is applied to all
>> hydrogens except the residues containing these soft core atoms.
>> To set up an equilibrated initial structure for the production TI
>> simulations, I was trying to minimize, heat and press the initial and the
>> end states.
>> After running a 50-ps(25000*0.002) heating step: From my MD output, I
>> found that the soft core atoms did not achieve a stable temperature of 310K.
>> Is it normal for soft core, non-SHAKE atoms to not achieve a stable
>> temperature? Or is it because we are calculating the temperature for only 3
>> atoms? Or is there anything I did wrong? :”(
>> Also, is there an upper limit for the number of soft core atoms? Does it
>> make more sense to include the whole residue? Will that drastically
>> increase the computing cost?
>> Any suggestion would be greatly appreciated. I look forward to learning
>> more about TI simulations.
>> Many Thanks,
>> Amy
>> --
>> Amy He
>> Chemistry Graduate Teaching Assistant
>> Hadad Research Group
>> Ohio State University
>> _______________________________________________
>> AMBER mailing list
Darrin M. York                    : Distinguished Professor
                                   : Henry Rutgers University Chair
Institute for Quantitative        :
    Biomedicine and Department of  : Director, Laboratory for
    Chemistry & Chemical Biology   :    Biomolecular Simulation Research
Rutgers, the State University     :
    of New Jersey                  :    fax:     +1-732-445-4320
174 Frelinghuysen Road            :    phone:   +1-848-445-5199
Piscataway, NJ 08854   USA        :;!!KGKeukY!2HVRmvKMxeBHFS3EoaMhud9k4msPmq5XFc_Innmcn4Kgm4wtK1Fc-C0hAHNQiFWorBed91caSLCsCrW2o4q-hINxYLM$<;!!KGKeukY!2HVRmvKMxeBHFS3EoaMhud9k4msPmq5XFc_Innmcn4Kgm4wtK1Fc-C0hAHNQiFWorBed91caSLCsCrW2o4q-hINxYLM$>
AMBER mailing list;!!KGKeukY!2HVRmvKMxeBHFS3EoaMhud9k4msPmq5XFc_Innmcn4Kgm4wtK1Fc-C0hAHNQiFWorBed91caSLCsCrW2o4q-51vN6mY$<;!!KGKeukY!2HVRmvKMxeBHFS3EoaMhud9k4msPmq5XFc_Innmcn4Kgm4wtK1Fc-C0hAHNQiFWorBed91caSLCsCrW2o4q-51vN6mY$>
AMBER mailing list
Received on Wed Jan 25 2023 - 16:00:03 PST
Custom Search