Re: [AMBER] Stability of a long NPT md trajectory?

From: Carlos Simmerling <carlos.simmerling.gmail.com>
Date: Mon, 9 Feb 2009 22:10:55 -0500

Bob,
I agree that random seed issues can be problematic, but I doubt it could
ever be anything of the scale seen here. his density is 3 times normal, but
for some reason the energy hasn't exploded and the pressure is too low! This
isn't just correlation issues, or non-ergodicity or something like that, it
is simply completely wrong at the most basic level. i bet it's the belly
option.
carlos


On Mon, Feb 9, 2009 at 9:21 PM, Robert Duke <rduke.email.unc.edu> wrote:

> I think this may be a really small unit cell (1000 waters - perhaps too
> small) with significant periodicity issues that then magnify the problems
> associated with usage of the same random seed through multiple restarts with
> ntt 3, but I am also somewhat skeptical that this is the only problem. For
> one thing, I did not get any indication of how many restarts were done with
> the same seed - may have been none. I mentioned it purely because it is
> something that folks tend to overlook, and it can cause really bad problems.
> Regards - Bob
>
> ----- Original Message ----- From: "Carlos Simmerling" <
> carlos.simmerling.gmail.com>
> To: <bill.mercury.chem.pitt.edu>; "AMBER Mailing List" <amber.ambermd.org>
> Sent: Monday, February 09, 2009 7:56 PM
>
> Subject: Re: [AMBER] Stability of a long NPT md trajectory?
>
>
> I disagree- I run frequently with ntt=3 and no ig value and have never
>> seen the problem you reported.
>>
>> are you sure that just changing ig and nothing else really solved it?
>>
>> beware of problems that go away without explanation- they often come
>> back without explanation too.
>>
>> as I said in my last email, if you use PME, make sure you understand
>> the limits of ibelly as compared to positional restraints, which think
>> are far safer. problems with ibelly and PME, especially during
>> equilibration, are well documented in the archives.
>>
>> On Mon, Feb 9, 2009 at 3:56 PM, Bill K <bill.mercury.chem.pitt.edu>
>> wrote:
>>
>>> Gentlemen,
>>> Thank you for the kind replies to my inquiry. I found your
>>> messages to be confusing initially until I realized my egregious
>>> mistake in applying the ibelly flag. I should have instead put in
>>> my input file (to freeze residue 1):
>>>
>>> Belly Residues (dynamic residues)
>>> RES 2 xxxx
>>> END
>>> END
>>> For edification purposes, though, I found that using ntt=3
>>> and not setting an ig value is what allowed the density to go
>>> such a high value.
>>> Thank you again for your time.
>>> -Bill
>>>
>>> There may be a number of things going on here relating to nonstandard
>>>> approach to md, but one thing I noticed immediately is the use of ntt =
>>>> 3.
>>>> This in itself is not a problem, but one must be certain to not use the
>>>> same
>>>> random seed at each restart - you must set a new value for "ig", and I
>>>> didn't notice it in the sample mdin. I don't understand at all why
>>>> your
>>>> density would be so high though - there is much more abnormal about this
>>>> run
>>>> than reuse of the same random seed in a restart for langevin dynamics.
>>>> Regards - Bob Duke
>>>>
>>>> ----- Original Message -----
>>>> From: "Carlos Simmerling" <carlos.simmerling.gmail.com>
>>>> To: <bill.mercury.chem.pitt.edu>; "AMBER Mailing List" <
>>>> amber.ambermd.org>
>>>> Sent: Monday, February 09, 2009 7:08 AM
>>>> Subject: Re: [AMBER] Stability of a long NPT md trajectory?
>>>>
>>>>
>>>> this should not happen.
>>>>> however, I suspect something related to your use of the belly option
>>>>> for nearly everything in the system. sander should have printed a
>>>>> warning about use of ibelly with pme- take a look at that. you may
>>>>> want to use resrtaints instead, but it's hard to give advice when we
>>>>> don't know why you defined only 1 dynamic residue.
>>>>>
>>>>>
>>>>> On Mon, Feb 9, 2009 at 2:07 AM, Bill K <bill.mercury.chem.pitt.edu>
>>>>> wrote:
>>>>>
>>>>>> Hi Everyone,
>>>>>> I'm running a relatively long (~10ns) trajectory of a single
>>>>>> solute in a water box of approx. 1000 tip3p waters, with standard
>>>>>> conditions of 300K and 1Atm. I apologize if this is an elementary
>>>>>> question, but I've been surfing the mail reflector for some time
>>>>>> now... I've brought the system to temperature
>>>>>> with NVT runs of at least 20ps (I've tried up to 1ns), and then
>>>>>> switched to NPT (using the input at the end of the message), with
>>>>>> ntt = 3. When I do this, I notice that, throughout the run, the
>>>>>> volume continues to decrease, and the pressure remains negative,
>>>>>> even after the system density surpasses 1 by a large margin.
>>>>>> the final output before the calculation crashes is:
>>>>>>
>>>>>> NMR restraints: Bond = 0.000 Angle = 0.000 Torsion =
>>>>>> 0.000
>>>>>>
>>>>>> NSTEP = 313500 TIME(PS) = 353.500 TEMP(K) = 569.51 PRESS
>>>>>> = -3544.4
>>>>>> Etot = -71984.1552 EKtot = 1.6976 EPtot
>>>>>> = -71985.8528
>>>>>> BOND = 0.0000 ANGLE = 0.0000 DIHED =
>>>>>> 0.0000
>>>>>> 1-4 NB = 0.0000 1-4 EEL = 0.0000 VDWAALS
>>>>>> -305.4604
>>>>>> EELEC = -71680.3923 EHBOND = 0.0000 RESTRAINT =
>>>>>> 0.0000
>>>>>> EKCMT = 0.2890 VIRIAL = 636.5491 VOLUME =
>>>>>> 8314.1509
>>>>>> Density =
>>>>>> 3.8796
>>>>>> Ewald error estimate: 0.9533E+00
>>>>>>
>>>>>>
>>>>>> So, instead of trying to do production runs using NTP, should I stop
>>>>>> the NTP run when my density is in the range of (1 or just over), and
>>>>>> then
>>>>>> carry on production runs at either NVT, or NVE with ntb=1 and ntt=0),
>>>>>> or
>>>>>> should this run be self regulating and stable and I've just completely
>>>>>> missed something? I appreciate your time and consideration.
>>>>>> thank you very much
>>>>>> -Bill
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> # 1ns Dynamic Simulation with Constant Pressure
>>>>>>
>>>>>> # Control section
>>>>>> &cntrl
>>>>>> ntwx = 500, ntpr = 500,
>>>>>> ntt = 3, gamma_ln = 1.0, temp0 = 300.0, tempi = 300.0, tautp = 1.0,
>>>>>> scnb = 2.0, scee = 1.2, dielc = 1, cut = 8.0,
>>>>>> ntb = 2, ntc = 2, ntf = 2,
>>>>>> nstlim = 999999, dt = 0.0010,
>>>>>> ntp = 1, taup = 5.0, pres0 = 1.0
>>>>>> ibelly = 1, ntr = 0,
>>>>>> imin = 0, irest = 0, ntx = 5, nmropt = 1,
>>>>>> /
>>>>>> &wt
>>>>>> type = 'TEMP0', istep1 = 1, istep2 = 999999, value1 = 300.0, value2 =
>>>>>> 300.0,
>>>>>> /
>>>>>> &wt
>>>>>> type='END'
>>>>>> /
>>>>>> Belly Residues (dynamic residues)
>>>>>> RES 1
>>>>>> END
>>>>>> END
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> 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
>>>>>
>>>>>
>>>>
>>>> _______________________________________________
>>>> 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
>>>
>>>
>> _______________________________________________
>> 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
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Feb 11 2009 - 01:13:59 PST
Custom Search