[gmx-users] Splitted DMPC bilayer

[Justin A. Lemkul]

Dr. Ramón Garduño-Juárez wrote:
> Justin,
> 
> Again much obliged for your comments. They are most illustrative...
> 
> I would like to make a final note on the issue of these many e-mails...
> 
> I am sure that GROMACS is fast, but that fast?...
> 

Yes.  Your results prove it.  With quality hardware, you get great performance.

> For the sake of knowing that we are doing the right things, this is our 
> topol.top file in which we eliminated all POSRES for the Protein and 
> DMPC, not so for the WATER...

To what end, I do not know.  One generally does not find much use in restraining 
water while everything else moves, but syntactically, it is correct.

> -------------
> ; Include forcefield parameters
> #include "./gromos53a6_lipid.ff/forcefield.itp"
> 
> ; Include chain topologies
> #include "topol_Protein_chain_A.itp"
> #include "topol_Protein_chain_B.itp"
> #include "topol_Protein_chain_C.itp"
> 
> ; Include water topology
> #include "./gromos53a6_lipid.ff/spc.itp"
> 
> #ifdef POSRES_WATER
> ; Position restraint for each water oxygen
> [ position_restraints ]
> ;  i funct       fcx        fcy        fcz
>    1    1       1000       1000       1000
> #endif
> 
> ; Include topology for ions
> #include "./gromos53a6_lipid.ff/ions.itp"
> 
> [ system ]
> ; Name
> mod.pdb
> 
> [ molecules ]
> ; Compound        #mols
> Protein_chain_A     1
> Protein_chain_B     1
> Protein_chain_C     1
> ----------------------
> 
> On Protein_chain_A  there are 342 atoms
> On Protein_chain_B  there are 289 atoms
> On Protein_chain_C  there are 715 atoms
> On DMPC there are 123 molecules of 46 atoms each
> On SOL there are 3205 molecules of 3 atoms each
> For a total of 16619 atoms
> 
> I know that this is a medium size system for which I was expecting 
> longer CPU time for a 20 ns MD run.
> 
> I know that there was no "error", which I meant is that I was surprised 
> by the outcome...
> 
> May be GROMACS is as fast as it is claimed...
> 

Indeed.

-Justin

> Cheers,
> Ramon
> 
> El 04/04/2011 05:27 p.m., Justin A. Lemkul escribió:
>>
>>
>> Dr. Ramón Garduño-Juárez wrote:
>>> Justin,
>>>
>>> Thank you for your comments after finishing the MD production run for 
>>> up to 20 ns...
>>>
>>> Since this step was over very quickly, now I have a simple question  
>>> ¿How long, in human time, should a production run last?
>>>
>>
>> There is no way to answer that.  It depends on the hardware, number of 
>> atoms, system load, application of any number of the Gromacs 
>> algorithms, .mdp settings...
>>
>>> The production run was carried out in six processors Intel Xeon (R) 
>>> E5405 2.00 GHz. The last few lines of the md_0_1.log are:
>>>
>>> -----------------------------------------
>>>         Parallel run - timing based on wallclock.
>>>
>>>                NODE (s)   Real (s)      (%)
>>>        Time: 180685.417 180685.417    100.0
>>>                        2d02h11:25
>>>                (Mnbf/s)   (GFlops)   (ns/day)  (hour/ns)
>>> Performance:    232.900     12.351      9.564      2.510
>>> -----------------------------------------
>>>
>>> Is this correct?   In my opinion it should lasted much more longer...
>>>
>>
>> Nope, Gromacs is just fast :)
>>
>>> Before reaching this point, this is an update of what we did...
>>>
>>> First we eliminated the SOL_SOL group and the only special index 
>>> group was Protein_DMPC.
>>>
>>> Since the NVT equilibration failed, we took option # 2 of the 
>>> "Advanced Troubleshooting", for the 1st phase of Equilibration.
>>>
>>> After this step we proceeded with the equilibration phase 2 with a 
>>> 1-ns NPT equilibration which ended fine.
>>>
>>> Next, we proceeded with a 20 ns production run. Thus, the modified 
>>> lines of the .mpd file found in the tutorial page were:
>>>
>>> nsteps           =  10000000  ;  2 * 10000000  =  2000  ps   (20 ns)
>>> tc-grps          =  Protein DMPC  SOL
>>> comm-grps  =  Protein_DMPC  SOL
>>>
>>> With this instructions the 20 ns simulation took  2d02h11:25
>>>
>>> I believe the "error" comes from the line
>>>
>>> constrains  =  all-bonds               which surely must be changed to
>>>
>>> constrains  =  none     or      hbonds
>>>
>>
>> Why do you say that?  What error is occurring?  You said your 
>> simulations were running fine.  You most certainly should not remove 
>> constraints if you're sticking with a 2-fs timestep.  The system will 
>> be unstable without constraints.  You might be able to get away with 
>> hbonds, but certainly not "none."
>>
>> -Justin
>>
>>> Looking forward to your comments...
>>>
>>> Much obliged,
>>> Ramon
>>>
>>

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================