[gmx-users] Correct .mdp file for Amber99SB-ILDN ff implementation in Gromacs
Justin Lemkul
jalemkul at vt.edu
Thu Jul 14 19:35:20 CEST 2016
On 7/14/16 12:40 PM, Mohsen Ramezanpour wrote:
> Thanks Joao and Justin for your comments,
>
> As Justin mentioned, I followed the original articles on this force field.
> However, I was not sure if the same parameters can be transferred to
> Gromacs or not.
>
The cutoffs are straightforward to get right. Not every software has the same
features, but the critical ones are all there.
> The only thing different was the initial temperature of 50K and increasing
> it to 310 K through simulation. I agree with you on this. But I had the
> following reason in mind. Please correct me if I am wrong.
> Since 310 is a normal T, 50 should change to 310, not a big difference.
That's a big difference.
> But, what do you think if you want to have 400 or 450 K as the equilibrium
> T? In this case, is not it better to just start with a smaller number, say
> 300 K, and raise it to 450 K? This is just an example to justify what I
> meant.
>
If you want to progressively change the temperature (for whatever reason), you
should do it properly via simulated annealing. If you start with, e.g. 50 K,
and tell your thermostat to regulate the temperature at 300 K, there's going to
be large scaling factors applied right at the outset of your simulation. That's
highly artificial and I would be skeptical of the integrity of that approach.
Either generate velocities at the corresponding temperature, or warm via
simulated annealing.
-Justin
> Cheers
> Mohsen
>
> On Thu, Jul 14, 2016 at 4:23 AM, Justin Lemkul <jalemkul at vt.edu> wrote:
>
>>
>>
>> On 7/14/16 4:37 AM, João Henriques wrote:
>>
>>> I was wondering if I should modify any terms because of using Gromacs
>>>>
>>> package?
>>>
>>> It's good practice to set your key parameters as close as possible to
>>> theirs. I am not familiar with Desmond, but I assume many of its
>>> parameters
>>> have similar counterparts in Gromacs. If you plan on (significantly)
>>> modifying any parameters, you necessarily need to perform a sanity check
>>> and validate your results.
>>>
>>> To me, and this is my personal opinion, there are some parameters in your
>>> email that seem a bit strange. Why 1 femtosecond integration time for the
>>> equilibration? 2 fs with LINCS should work fine. Why do you set two
>>> different Fourier spacings? Plus, why set the last one (the one that is
>>>
>>
>> The OP will undoubtedly get a fatal error from repeating fourierspacing.
>> This comes from too literally following the paper's methods. The authors
>> in that case specified 32 grid points in each spatial dimension, but this
>> is a system-dependent quantity, and for a large system this may be too
>> inaccurate. The OP should probably leave the default fourierspacing (0.12)
>> alone and not set fourier-n[xyz] explicitly.
>>
>> effectively used) to zero? Where did you get that 1.05 for the rlist? Is
>>> that 0.05 going to make any difference? Also, double check the pressure
>>> and
>>>
>>
>> This is following the paper, in which they state it explicitly as rlist.
>> I agree that such a short distance is unlikely to have a practical effect,
>> but in the effort to follow the force field's validated run parameters, it
>> is correct.
>>
>> In GROMACS, using the Verlet cutoff scheme with rlist = rcoulomb = rvdw =
>> 1.0 would be a faithful representation of what the AMBER99SB-ILDN paper did.
>>
>> temperature coupling constants. Finally, why are you generating an initial
>>> temperature of 50K and then have it coupled to 310 K?
>>>
>>> These are just a few issues that come to mind after a quick read of your
>>> email. There may be others, and I may also be wrong in my assessment... I
>>> personally believe that there is no single recipe to performing
>>> simulations, and I find that it is always better to modify only what you
>>> need and understand, instead of following someone's work blindly. After
>>> all, you need you understand what and why you do it and, unless you're
>>> trying to replicate their work, no single set of parameters should be
>>> absolutely "right". You need, of course, to be sensible when changing them
>>> and check/validate your results afterwards.
>>>
>>>
>> Agreed, except in the case of cutoffs and water models. These are as much
>> a part of the force field as the parameters themselves. The effects due to
>> mistakes can be subtle and unpredictable, so unless the user is prepared to
>> do a careful, systematic study of cutoff effects (which is probably a paper
>> or two in and of itself), then that user should follow the people who
>> actually made the force field. A model is only valid under the conditions
>> which are known to be valid :)
>>
>> -Justin
>>
>>
>> Hope it helps,
>>> Best regards,
>>> João
>>>
>>> On Wed, Jul 13, 2016 at 11:34 PM, Mohsen Ramezanpour <
>>> ramezanpour.mohsen at gmail.com> wrote:
>>>
>>> Dear Gromacs-users,
>>>>
>>>> I am interested in simulation of protein-ligand system.
>>>>
>>>> To do so, I have chosen the amber99sb-ildn force field.
>>>>
>>>> Reading through the original paper on this force filed by D.E.Shaw group
>>>> (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2970904/) I have made the
>>>> following equilibriu,.mdp and production-run.mdp files. I am not sure of
>>>> it, though. Please correct me if I am wrong.
>>>>
>>>> The authors have used Desmond package not Gromacs for their simulations.
>>>> I was wondering if I should modify any terms because of using Gromacs
>>>> package?
>>>> (This is the case for using charmm force field in Gromacs).
>>>>
>>>> I read other posts in gromacs list but I could not get any clear answer
>>>> for
>>>> this.
>>>>
>>>> Thanks in advance for your comments
>>>> Cheers
>>>> Mohsen
>>>>
>>>>
>>>> for Equilibrium.mdp, I made:
>>>>
>>>> define =
>>>>
>>>> ; RUN CONTROL PARAMETERS
>>>> integrator = md ;
>>>> tinit = 0 ;
>>>> dt = 0.001 ; 1 femtosecond time step for
>>>> integration
>>>> nsteps = 500000 ; 1 ns
>>>>
>>>> ; OUTPUT CONTROL OPTIONS
>>>> nstxout = 0 ; Writing full precision
>>>> coordinates
>>>> nstvout = 0 ; Writing velocities
>>>> nstfout = 0 ; writing forces
>>>> nstlog = 2500 ; Writing to the log file
>>>> nstenergy = 2500 ; Writing out energy information
>>>> nstxtcout = 2500 ; Writing coordinates every 5 ps
>>>> energygrps = Protein ligand Water_and_ions
>>>> xtc_precision = 1000
>>>> xtc-grps = System
>>>>
>>>> ; NEIGHBORSEARCHING PARAMETERS
>>>> nstlist = 10
>>>> ns-type = Grid
>>>> pbc = xyz
>>>> rlist = 1.05
>>>> cutoff-scheme = Verlet
>>>> nstcalclr = 1
>>>>
>>>> ; OPTIONS FOR ELECTROSTATICS AND VDW
>>>> coulombtype = PME
>>>> rcoulomb = 1.0
>>>> fourierspacing = 0.1
>>>> vdw-type = cut-off
>>>> rvdw = 1.0
>>>> optimize_fft = yes
>>>> fourierspacing = 0
>>>> fourier-nx = 32
>>>> fourier-ny = 32
>>>> fourier-nz = 32
>>>> pme-order = 4
>>>>
>>>> ; Temperature coupling
>>>> Tcoupl = v-rescale ; in the original paper
>>>> it
>>>> is Nose-Hoover thermostat
>>>> tc-grps = Protein_ligand Water_and_ions
>>>> tau_t = 1.0 1.0
>>>> ref_t = 310 310
>>>>
>>>> ; Pressure coupling
>>>> Pcoupl = nose-hoover
>>>> Pcoupltype = isotropic
>>>> tau_p = 5.5
>>>> compressibility = 4.5e-5
>>>> ref_p = 1.00
>>>> refcoord_scaling = com
>>>>
>>>> ; GENERATE VELOCITIES FOR STARTUP RUN
>>>> gen_vel = yes ; Assign velocities to particles by
>>>> taking them randomly from a Maxwell distribution
>>>> gen_temp = 50.0 ; Temperature to generate corresponding
>>>> Maxwell distribution
>>>> gen_seed = 1235 ; Seed for (semi) random number
>>>> generation.
>>>>
>>>> ; OPTIONS FOR BONDS
>>>> constraint-algorithm = SHAKE
>>>> constraints = h-bonds
>>>>
>>>>
>>>> and for production-run.mdp, I just change the
>>>>
>>>> dt = 0.002 ; 1 femtosecond time step
>>>> for
>>>> integration
>>>> constraint-algorithm = LINCS
>>>> constraints = all-bonds
>>>> gen_vel = no
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> *Rewards work better than punishment ...*
>>>> --
>>>> Gromacs Users mailing list
>>>>
>>>> * Please search the archive at
>>>> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
>>>> posting!
>>>>
>>>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>>>
>>>> * For (un)subscribe requests visit
>>>> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
>>>> send a mail to gmx-users-request at gromacs.org.
>>>>
>>>>
>> --
>> ==================================================
>>
>> Justin A. Lemkul, Ph.D.
>> Ruth L. Kirschstein NRSA Postdoctoral Fellow
>>
>> Department of Pharmaceutical Sciences
>> School of Pharmacy
>> Health Sciences Facility II, Room 629
>> University of Maryland, Baltimore
>> 20 Penn St.
>> Baltimore, MD 21201
>>
>> jalemkul at outerbanks.umaryland.edu | (410) 706-7441
>> http://mackerell.umaryland.edu/~jalemkul
>>
>> ==================================================
>>
>> --
>> Gromacs Users mailing list
>>
>> * Please search the archive at
>> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
>> posting!
>>
>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>
>> * For (un)subscribe requests visit
>> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
>> send a mail to gmx-users-request at gromacs.org.
>>
>
>
>
--
==================================================
Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow
Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201
jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul
==================================================
More information about the gromacs.org_gmx-users
mailing list