[gmx-users] Coarse-grained Protein-ligand simulations

Tue Apr 2 10:54:24 CEST 2019

Hi Mac Kevin E. Braza,

Would it be possible to use a GPU?  If can manage with single precision 
and need less than 8Gb of memory, a gaming GPU might give some 
performance improvement.

Regards,

Benson

On 4/2/19 11:46 AM, Peter Kroon wrote:
> @Joao: I didn't mean to imply in any way that everyone has (or should
> have) a couple hundred nodes at their beck and call. Although it would
> be nice. I underestimated the size of the GPCR complex, as well as how
> slow atomistic simulations are :)
>
> Now that I've tried to pull my foot out of my mouth again, back on
> topic: although Martini can do (almost) anything, I am rather skeptical
> of CG docking/binding studies because of the reasons I outlined earlier.
> In addition, in Martini the error in the entropic term (due to a lack of
> conformational freedom) in the free energy equation is compensated in
> the enthalpic term. However, in a confined environment (protein pocket!)
> this compensation may have to be different --- and the ligand was
> parametrised in solution.
>
>
> Peter
>
>
> On 02-04-19 05:18, Billy Williams-Noonan wrote:
>> Have you considered accelerated MD?  Like metadynamics.  Plumed has a lot
>> of options there
>>
>> Cheers,
>> Billy
>>
>> On Tue, 2 Apr 2019 at 09:18, Mac Kevin Braza <mebraza at up.edu.ph> wrote:
>>
>>> Hello Sir Benson,
>>>
>>> We are using Supermicro SYS-1028R-WC1R Server with 2 x 2.2Ghz 12-Core Intel
>>> Processors
>>> (4 x 8GB DDR4) with a single node only. Ideally, to reach the microsecond
>>> simulation of GPCR-membrane
>>> simulation in all-atom, we will be needing a computer cluster with at least
>>> 200 parallel nodes system.
>>> But even with a 50-100 parallel nodes, we will reach the simulation time
>>> for a month, although we know that this is
>>> challenging for us here in the Philippines.
>>>
>>> The specialized super-computer cluster Anton is an example of hardware that
>>> have reached more than 100 microseconds
>>> simulation of the all-atom GPCR-membrane simulation in a month of total CPU
>>> time. It has 512 processing nodes.
>>>
>>> Best regards,
>>> Mac Kevin E. Braza
>>>
>>> On Tue, Apr 2, 2019, 12:40 AM Benson Muite <benson_muite at emailplus.org>
>>> wrote:
>>>
>>>> Hi Mac Kevin E. Braza,
>>>>
>>>> What hardware are you using? What kind of hardware would be needed to do
>>>> a full simulation instead of a coarse-grained one?
>>>>
>>>> Regards,
>>>>
>>>> Benson
>>>>
>>>> On 4/1/19 6:49 PM, João Henriques wrote:
>>>>> GPCR + membrane systems are notoriously big systems to work with for
>>> most
>>>>> research groups, regardless of your location on the map. Even in
>>>>> "privileged Europe" many research groups would struggle to produce
>>>>> microsecond long atomistic simulations of this system within a short
>>>> period
>>>>> of time. Moreover, "privileged Europe" is also home to significant
>>>> computer
>>>>> resource discrepancies among its member countries. This is actually one
>>>> of
>>>>> the main reasons why your group's CG model is so popular :)
>>>>>
>>>>> On Mon, Apr 1, 2019 at 5:09 PM P C Kroon <p.c.kroon at rug.nl> wrote:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> I work in privileged Europe, so it’s good for me to get a reality
>>> check
>>>>>> once every while. Thanks.
>>>>>>
>>>>>> Coarse graining molecules for Martini is not too hard. There should be
>>>>>> some tutorials on cgmartini.nl that should help you get underway. You
>>>>>> will, however, run into the problems I mentioned, and you will need to
>>>> do
>>>>>> extensive validation on the topologies of your ligands. Again, it
>>>> depends
>>>>>> on your exact research question: if you’re doing high-throughput like
>>>>>> screening, qualitative models might be good enough. Also see T
>>> Bereau’s
>>>>>> automartini.
>>>>>>
>>>>>> Peter
>>>>>>
>>>>>> From: Mac Kevin Braza
>>>>>> Sent: 01 April 2019 16:06
>>>>>> To: gmx-users at gromacs.org
>>>>>> Cc: gromacs.org_gmx-users at maillist.sys.kth.se
>>>>>> Subject: Re: [gmx-users] Coarse-grained Protein-ligand simulations
>>>>>>
>>>>>> Dear Sir Peter Kroon,
>>>>>>
>>>>>> We are currently maximizing the computer capabilities to reach
>>>> microsecond,
>>>>>> but to reach 1 microsecond in our lab, it would take me at least 6
>>>> months
>>>>>> to finish all one microsecond.
>>>>>> We do not have that high level capacities here in the Philippines to
>>>> reach
>>>>>> it. Membrane proteins are
>>>>>> typically longer, with all the lipid bilayers, solvent, and ions
>>>> present on
>>>>>> top of the protein.
>>>>>> We will need more powerful computers in this part.
>>>>>>
>>>>>> I found few works from literature on the protein-ligand representation
>>>> in
>>>>>> Coarse-grained.
>>>>>> We found several papers but they are either have vague methodology in
>>>>>> describing the ligand coarse-graining method and/or not necessarily
>>> have
>>>>>> the same research problem
>>>>>> as we want to explore.
>>>>>>
>>>>>> All in all, we will finish the simulation in all-atom as long as we
>>> can,
>>>>>> and still be hopeful with
>>>>>> the coarse-graining method. What we explored as in the present is the
>>>>>> CHARMM-GUI Martini Maker,
>>>>>> yet they do not include the drug ligands in representing them in
>>>>>> coarse-grained. I still have to search for other means
>>>>>> to do this. Thank you very much!
>>>>>>
>>>>>> Best regards,
>>>>>> Mac Kevin E. Braza
>>>>>>
>>>>>> On Mon, Apr 1, 2019 at 5:59 PM Peter Kroon <p.c.kroon at rug.nl> wrote:
>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> that's probably a tough cookie. My first instinct would be to just
>>>> apply
>>>>>>> a more hardware, and do it all atomistically. A microsecond should be
>>>>>>> within reach. Whether it's enough is a separate matter. The problem
>>> is
>>>>>>> that most CG representations don't get the shape of both your pocket
>>>> and
>>>>>>> ligand exactly right, producing unreliable answers. In addition, in
>>>> most
>>>>>>> CG FFs hydrogen bonds are isotropic and not specific enough for this
>>>>>>> kind of problem.
>>>>>>>
>>>>>>> If "more hardware" is not an option you'll need to dive into
>>> literature
>>>>>>> to see if people did CG protein-ligand binding/docking/unbinding
>>>>>>> (depening on research question). I would also be very skeptical of
>>> any
>>>>>>> (absolute) kinetics produced by CG simulations.
>>>>>>>
>>>>>>> As a last ditch effort you could look into multiscaling, but that's a
>>>>>>> research topic in its own.
>>>>>>>
>>>>>>>
>>>>>>> Peter
>>>>>>>
>>>>>>>
>>>>>>> On 01-04-19 11:49, Mac Kevin Braza wrote:
>>>>>>>> Thank you Prof. Lemkul,
>>>>>>>>
>>>>>>>> I appreciate your comment on this part.
>>>>>>>>
>>>>>>>> Sir Peter Kroon,
>>>>>>>>
>>>>>>>> We want to do the coarse-grained MD simulation to access long
>>>> timescale
>>>>>>>> events of the
>>>>>>>> effect of the ligand binding to the GPCR, at least microsecond . For
>>>>>> now,
>>>>>>>> the most accessible means for us is to
>>>>>>>> do the CGMD. But we are currently being cornered in choosing which
>>>>>> set-up
>>>>>>>> will best suit, and
>>>>>>>> if it will allow us to see these events. We are looking also in the
>>>>>>>> possibility of coarse-graining
>>>>>>>> the ligand, and if you can share your expertise in coarse-graining
>>>> also
>>>>>>> the
>>>>>>>> ligand that would be great.
>>>>>>>> I appreciate this Sir Kroon, thank you very much!
>>>>>>>>
>>>>>>>> Best regards,
>>>>>>>> Mac Kevin E. Braza
>>>>>>>>
>>>>>>>> On Mon, Apr 1, 2019 at 5:07 PM Peter Kroon <p.c.kroon at rug.nl>
>>> wrote:
>>>>>>>>> If I may chip in: It really depends on what you're studying, and
>>> what
>>>>>>>>> forcefield you're using to do it. Unfortunately there is no FF that
>>>>>>>>> reproduces all behaviour accurately. The art is in picking one that
>>>>>> (at
>>>>>>>>> least) reproduces what you're interested in.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Peter
>>>>>>>>>
>>>>>>>>> On 29-03-19 17:26, Justin Lemkul wrote:
>>>>>>>>>> On 3/29/19 9:17 AM, Mac Kevin Braza wrote:
>>>>>>>>>>> Thank you Professor Lemkul,
>>>>>>>>>>>
>>>>>>>>>>> But would you suggest on how can I coarse-grained the ligand I am
>>>>>>>>>>> using? I
>>>>>>>>>>> have been searching resources online but they do not work in our
>>>>>> part.
>>>>>>>>>> I don't work with CG simulations, so I'm not much help. I would
>>>> think
>>>>>>>>>> that a CG parametrization of a ligand would remove all the detail
>>>>>>>>>> you'd normally want to see in terms of ligand-protein
>>> interactions.
>>>>>>>>>> -Justin
>>>>>>>>>>
>>>>>>>>>>> I hope you can help us. Thank you Prof. Lemkul!
>>>>>>>>>>>
>>>>>>>>>>> Best regards,
>>>>>>>>>>> Mac Kevin E. Braza
>>>>>>>>>>>
>>>>>>>>>>> On Fri, Mar 29, 2019, 8:59 PM Justin Lemkul <jalemkul at vt.edu>
>>>>>> wrote:
>>>>>>>>>>>> On 3/29/19 3:32 AM, Mac Kevin Braza wrote:
>>>>>>>>>>>>> Hello everyone,
>>>>>>>>>>>>>
>>>>>>>>>>>>> I am simulating a coarse-grained model of a membrane protein
>>>>>> (GPCR)
>>>>>>> in
>>>>>>>>>>>>> lipid bilayer and an all-atom ligand octopamine. I build the
>>>>>>> protein,
>>>>>>>>>>>>> solutes, and membrane in the web server CHARMM-GUI. While, I
>>>> added
>>>>>>> the
>>>>>>>>>>>>> ligand to the protein complex manually using the same
>>> coordinates
>>>>>>>>>>>>> of the
>>>>>>>>>>>>> coarse-grained protein model.
>>>>>>>>>>>>>
>>>>>>>>>>>>> I used the GROMACS input files from the output of CHARMM-GUI to
>>>>>>>>>>>>> simulate
>>>>>>>>>>>>> the system. I include the LIGAND.ITP (from the PRODRG Server)
>>> to
>>>>>> the
>>>>>>>>>>>>> system.top and added the atom indexes in the index.ndx file.
>>>>>>>>>>>> Don't do this. An atomistic representation of a ligand and a CG
>>>>>>>>>>>> representation of everything else is incompatible. Mixing and
>>>>>>> matching
>>>>>>>>>>>> force fields is never a good idea. Moreover, PRODRG produces
>>>>>>> topologies
>>>>>>>>>>>> that are known to be unsuitable for MD simulations.
>>>>>>>>>>>>
>>>>>>>>>>>>> However, when I proceed with the second part of equilibration,
>>>> the
>>>>>>>>>>>>> following errors occurred.
>>>>>>>>>>>>>
>>>>>>>>>>>>> *Command line*:
>>>>>>>>>>>>>       gmx grompp -f step6.2_equilibration.mdp -o
>>>>>>>>>>>>> step6.2_equilibration.tpr
>>>>>>>>>>>> -c
>>>>>>>>>>>>> step6.1_equilibration.gro -p system.top -n index.ndx
>>>>>>>>>>>>>
>>>>>>>>>>>>> Setting the LD random seed to 1722366284
>>>>>>>>>>>>> Generated 2391 of the 4656 non-bonded parameter combinations
>>>>>>>>>>>>> Excluding 1 bonded neighbours molecule type 'PROA_P'
>>>>>>>>>>>>> Excluding 1 bonded neighbours molecule type 'POPC'
>>>>>>>>>>>>> Excluding 1 bonded neighbours molecule type 'W'
>>>>>>>>>>>>> Excluding 1 bonded neighbours molecule type 'NA'
>>>>>>>>>>>>> Excluding 1 bonded neighbours molecule type 'CL'
>>>>>>>>>>>>> Excluding 3 bonded neighbours molecule type 'LIG'
>>>>>>>>>>>>> Velocities were taken from a Maxwell distribution at 303.15 K
>>>>>>>>>>>>> Removing all charge groups because cutoff-scheme=Verlet
>>>>>>>>>>>>>
>>>>>>>>>>>>> -------------------------------------------------------
>>>>>>>>>>>>> Program gmx grompp, VERSION 5.1.4
>>>>>>>>>>>>> Source code file:
>>>>>>>>>>>>> /home/gromacs-5.1.4/src/gromacs/gmxpreprocess/readir.c,
>>>>>>>>>>>>> line: 2690
>>>>>>>>>>>>>
>>>>>>>>>>>>> Fatal error:
>>>>>>>>>>>>> 20 atoms are not part of any of the T-Coupling groups
>>>>>>>>>>>>> For more information and tips for troubleshooting, please check
>>>>>> the
>>>>>>>>>>>> GROMACS
>>>>>>>>>>>>> website at http://www.gromacs.org/Documentation/Errors
>>>>>>>>>>>>> -------------------------------------------------------
>>>>>>>>>>>>>
>>>>>>>>>>>>> The 20 atoms described the ligand I placed inside the
>>>>>>> protein-membrane
>>>>>>>>>>>>> complex. I want to know if where can this error originate and
>>> how
>>>>>>>>>>>>> can we
>>>>>>>>>>>>> fix them?
>>>>>>>>>>>> This simply means you haven't specified the ligand anywhere in
>>>>>>> tc-grps.
>>>>>>>>>>>> But again, back up and reevaluate your approach, which is far
>>> more
>>>>>>>>>>>> problematic than this simple index group issue.
>>>>>>>>>>>>
>>>>>>>>>>>> -Justin
>>>>>>>>>>>>
>>>>>>>>>>>> --
>>>>>>>>>>>> ==================================================
>>>>>>>>>>>>
>>>>>>>>>>>> Justin A. Lemkul, Ph.D.
>>>>>>>>>>>> Assistant Professor
>>>>>>>>>>>> Office: 301 Fralin Hall
>>>>>>>>>>>> Lab: 303 Engel Hall
>>>>>>>>>>>>
>>>>>>>>>>>> Virginia Tech Department of Biochemistry
>>>>>>>>>>>> 340 West Campus Dr.
>>>>>>>>>>>> Blacksburg, VA 24061
>>>>>>>>>>>>
>>>>>>>>>>>> jalemkul at vt.edu | (540) 231-3129
>>>>>>>>>>>> http://www.thelemkullab.com
>>>>>>>>>>>>
>>>>>>>>>>>> ==================================================
>>>>>>>>>>>>
>>>>>>>>>>>> --
>>>>>>>>>>>> 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.
>>>>>>>>>>>>
>>>>>>>>> --
>>>>>>>>> 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.
>>>>>>> --
>>>>>>> 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.
>>>>>>>
>>>>>> --
>>>>>> 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.
>>>>>>
>>>>>> --
>>>>>> 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.
>>>> --
>>>> 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.
>>> --
>>> 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.
>>