Hi David,
I used two steps of minimizations without shake. The first minimization is for solution and the second is for the whole system.
The script for the first minimization is the following:
$cntrl
imin = 1,! (0 for simulation without minimization)
maxcyc = 10000,
ncyc = 5000, !(after steps of steepest descent, switch to conjugate gradient)
ntmin = 1, !(0, full conjugate;1, steepest descent and conjugate;2,steepest descent;3, XMIN;4,LMOD)
ntpr=10, !(print energy frequency)
ntr=1, !(turn on Cartesian restraints)
restraint_wt=9999.0, !(force constant for restraint)
restraintmask=':1-448', !(atoms in residues 1-58 restrained)
ntwx = 100, !write crd file every ntwx steps
ntwe = 100, !write energy
ntwf = 100, !write force
ntwv = 100, !write velocity
ntwprt = 0, ! how many atoms to write to output crd and vel file,0 is for all atoms
idecomp = 0, !energy decomposition
/
The output for the first minimization:
NSTEP ENERGY RMS GMAX NAME NUMBER
5114 -1.5910E+05 2.9421E+01 5.8007E+03 HB2 1631
BOND = 6917.9314 ANGLE = 1092.4637 DIHED = 4959.0516
VDWAALS = 12825.8823 EEL = -215466.3252 HBOND = 0.0000
1-4 VDW = 8962.8025 1-4 EEL = 19879.4322 RESTRAINT = 1724.6259
EAMBER = -160828.7615
***** REPEATED LINMIN FAILURE *****
The script for the second minimization is the following:
$cntrl
imin = 1,! (0 for simulation without minimization)
maxcyc = 10000,
ntpr = 100,
ncyc = 5000, !(after steps of steepest descent, switch to conjugate gradient)
ntmin = 1, !(0, full conjugate;1, steepest descent and conjugate;2,steepest descent;3, XMIN;4,LMOD)
ntwx = 100, !write crd file every ntwx steps
ntwe = 100, !write energy
ntwf = 100, !write force
ntwv = 100, !write velocity
ntwprt = 0, ! how many atoms to write to output crd and vel file,0 is for all atoms
idecomp = 0, !energy decomposition
/
I just noticed the linmin failure in the first minimization.
What is the reasonable region for RMS to tell good minimization?
I also notice some of my systems have large RMS from the first minimization (RMS ~8) but low second minimization (RMS ~0.5), does that mean I still need to increase the steps for the first minimization?
BTW, do you mean first heat the system with solute fixed and then relax the system or fixing the solute in the minimization process?
Thanks,
Zizhang
> On Jan 10, 2020, at 3:54 PM, David Case <david.case.rutgers.edu> wrote:
>
> On Fri, Jan 10, 2020, zizhang sheng wrote:
>>
>> Energy minimization:
>> NSTEP ENERGY RMS GMAX NAME NUMBER
>> 5214 -1.5853E+05 4.6706E+01 8.2567E+03 HB2 1631
>>
>> BOND = 6923.8637 ANGLE = 1092.5932 DIHED = 4959.0141
>> VDWAALS = 12825.5765 EEL = -215473.5524 HBOND = 0.0000
>> 1-4 VDW = 11265.8123 1-4 EEL = 19879.2193 RESTRAINT = 0.0000
>>
>> Heating:
>>
>> NSTEP = 0 TIME(PS) = 0.000 TEMP(K) = 0.00 PRESS = 0.0
>> Etot = -164877.0775 EKtot = 0.0000 EPtot = -164877.0775
>> BOND = 624.2695 ANGLE = 1092.5932 DIHED = 4959.0141
>> 1-4 NB = 11265.8123 1-4 EEL = 19879.2193 VDWAALS = 12761.0383
>> EELEC = -215459.0242 EHBOND = 0.0000 RESTRAINT = 0.0000
>> ------------------------------------------------------------------------------
>>
>> NMR restraints: Bond = 0.000 Angle = 0.000 Torsion = 0.000
>>
>> It seems the bond energy change dramatically between the two files.
>
> Check how shake is set in minimization vs. MD.
>
>>
>> I am now running the same heating script with vlimit=10. For both
>> PMEMD.MPI and PMEMD.CUDA, the heating step finished ok. But the output
>> shows lots of velocity exceeding 10, even in thousands. I wonder if this
>> will help find the cause.
>
> You probably need to either minimize better (you still have a large RMS
> gradient at the end of minimization); or (more likely to work) heat much
> more slowly: set temp0 to 100., dt=0.001, gamma_ln=10. and run a few
> thousand steps of MD. I bet you'll see your potential energy go down a
> lot, and (one hopes) stabilize around a new value. After that, you can
> probably heat more normally.
>
> Basically, if you still trying to do all your heating in a single run
> with 2.5 million steps, that is the wrong approach. You may need to
> carry out shorter MD runs with greater temperature control until the
> system gets stabilized. Play with MD parameters for runs with 2.5
> thousand (not 2.5 million) steps (or maybe even just 250 steps). Every
> system is different, and you will need to experiment. I also notice
> that you have no restraints defined -- you should consider adding
> restaints to keep the solute fixed near its starting configuration until
> you have done lot of MD equilibration, then slowly relaxing those
> constraints. Otherwise, bad forces at the beginning may distort your
> structure and it might never find its way back.
>
> ....dac
>
>> mG
>> Best,
>>
>> Zizhang
>>
>>
>>> On Jan 9, 2020, at 9:18 AM, zizhang sheng <shengzizhang.gmail.com> wrote:
>>>
>>> Hi All,
>>>
>>> When I run MD for glycoproteins in Amber18 pmemed.cuda.MPI, it always crash with the memory error when heating from 0k to 300k. The simulations without glycans but with identical MD parameters are all good. The glycan I added is MAN3 from Glycam. Tleap did not show any error. I also tried with the vlimit=20 option suggested in previous threads but run into the same error. I also tried with pmem.cuda and pmem.MPI, they all give the same error. The GPU card is 1080Ti. I wonder if anyone has the similar issue when simulating glycoprotein.
>>>
>>> The steps I am running are:
>>> 1) minimize the solution
>>> 2) minimize the whole system
>>> 3) heating
>>> 4) equilibration
>>>
>>> Attached is the parameters for heating.
>>>
>>> Thanks,
>>>
>>> Zizhang Sheng
>>>
>>>
>>> &cntrl
>>> nstlim=2500000, dt=0.002, ntx=1, irest=0, ntpr=20000, ntwr=20000, ntwx=20000,
>>> tempi =0.0, temp0=300.0, ntt=3, gamma_ln=1.0,taup=5,
>>> vlimit=20,
>>> ntb=1, ntp=0,
>>> cut=12,
>>> ntc=2, ntf=2,
>>> nmropt=1,
>>> iwrap=1,
>>> nrespa=1,!frequency of steps to evaluate slowly-varying terms
>>> /
>>>
>>> &wt type='TEMP0', istep1=0, istep2=900000, value1=0.0, value2=300.0 /
>>> &wt type='TEMP0', istep1=900001, istep2=2500000, value1=300.0, value2=300.0 /
>>> &wt type='END’ /
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> https://nam02.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.ambermd.org%2Fmailman%2Flistinfo%2Famber&data=02%7C01%7Cdavid.case%40rutgers.edu%7C3de3bc98b02244188e4108d795f2c29a%7Cb92d2b234d35447093ff69aca6632ffe%7C1%7C1%7C637142742190645199&sdata=%2FDwHTDGMIWvz0qD5aby5HtE%2F%2BunOVI5icaS5o3F98DY%3D&reserved=0
>
> --
>
> ====================================================================
> David A. Case | david.case.rutgers.edu
> Dept. of Chemistry & Chemical Biology |
> Rutgers University | office: +1-848-445-5885
> 174 Frelinghuysen Road, Rm. 208b | cell: +1-609-751-8668
> Piscataway, NJ 08854 USA | web: casegroup.rutgers.edu
> ====================================================================
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Jan 10 2020 - 14:00:02 PST
