Re: Rigid bonds and non-bonded energy

From: Grzegorz Nawrocki (aksonik_at_gmail.com)
Date: Wed Oct 21 2015 - 08:29:22 CDT

Yes, except rigid bonds all the settings are the same, and they are the
simplest as possible, i.e. no PBC, no cutoff (very large), no switching, no
PME, etc. It runs on single CPU.

2015-10-20 13:52 GMT-04:00 Brian Radak <brian.radak.accts_at_gmail.com>:

> Interesting, so these are systematic energy differences on the order of
> 0.1 kcal/mol; that does seem concerning. I assume you are using PME? The
> same FFT settings? Does the agreement change/improve if you run on a single
> core?
>
>
> On 10/20/2015 11:09 AM, Grzegorz Nawrocki wrote:
>
> I play with a small box of water, i.e. about 2.5 nm edge. Generate an
> output structure with rigid bonds and use it as an input. Repeat this until
> input and output are exactly the same. The electrostatic and VDW energies
> are equal to -15731 and 29873.3 [kJ/mol], respectively. Then I turn off
> rigid bonds (input and output structures are always the same) and the
> energies are -15729.9 and 29872.9. This kind of difference I am talking
> about.
>
> 2015-10-20 11:33 GMT-04:00 Brian Radak <brian.radak.accts_at_gmail.com>:
>
>> "Much more similar" still sounds a bit concerning to me. I can generally
>> get matching to all precision given by NAMD (i.e 1e-4), although frequently
>> that is for single molecules in vacuum. Perhaps the variability of multiple
>> threads on a large structure could plausibly change things, although even
>> then variation at 1e-3 is a bit concerning.
>>
>> The only time I've noticed trouble is that when reading in velocities the
>> kinetic energy tends to drift by about ~0.1 kcal/mol (although turning
>> on/off constraints will also affect this due to changes in degrees of
>> freedom); in any event, you are not doing this.
>>
>> Brian
>>
>>
>> On 10/20/2015 10:11 AM, Grzegorz Nawrocki wrote:
>>
>> Thank you Brian very much for the explanation. I hadn't known about
>> coordinates resetting and I was sure that analyse exactly the same
>> structure. Indeed, with rigid bonds output structure is slightly different
>> than input one. When I use the former as input and turn of rigid bonds the
>> non-bonded energies are much more similar.
>>
>> Best regards,
>>
>> 2015-10-19 15:40 GMT-04:00 Brian Radak < <brian.radak.accts_at_gmail.com>
>> brian.radak.accts_at_gmail.com>:
>>
>>> RATTLE is just the constraint algorithm used in NAMD (I believe just the
>>> original SHAKE algorithm plus some extra stuff to handle velocities). After
>>> RATTLE is applied the coordinates are reset such that they abide by the
>>> constraints. If the constraints are already satisfied, then this will have
>>> no effect.
>>>
>>> The obvious thing to do here would be to turn off rigidBonds and then
>>> see if the energies match. If they do, then you are probably seeing an
>>> effect from constraints. This happens to me a lot when I build protons with
>>> psfgen and they are either wrongly placed at the origin or given an
>>> non-ideal bond length.
>>>
>>>
>>>
>>> On Mon, Oct 19, 2015 at 9:22 AM, Grzegorz Nawrocki < <aksonik_at_gmail.com>
>>> aksonik_at_gmail.com> wrote:
>>>
>>>> Dear Brian,
>>>> Thank you for the reply. Yes, the bonds have optimal lengths.
>>>>
>>>> I am running dynamics, but the input and output structure is the same.
>>>> It means that coordinates don't change at all, that is confirmed by zero
>>>> kinetic energy. Non-bonded interactions depend only on positions, so I
>>>> should get exactly the same values. If the energy is neither of the input
>>>> nor output structure, what is the structure that energy comes from? Could
>>>> you please explain me what you mean by saying "rattle" or "coordinate
>>>> resetting"?
>>>>
>>>> 2015-10-16 18:42 GMT-04:00 Brian Radak < <brian.radak.accts_at_gmail.com>
>>>> brian.radak.accts_at_gmail.com>:
>>>>
>>>>> Is your structure generated in accord with those constraints? Since
>>>>> you are running dynamics, I believe RATTLE will have a go at coordinate
>>>>> resetting. In that case the energy you get will not be the energy of the
>>>>> input structure if the constraints are not satisfied.
>>>>>
>>>>> On 10/16/2015 04:39 PM, Grzegorz Nawrocki wrote:
>>>>>
>>>>>> Dear NAMD users,
>>>>>> I run simulation with number of steps equal to zero, since I want to
>>>>>> calculate energy of a single structure. Log file says that kinetic energy
>>>>>> is zero, that is fine, but non-bonded energies, both VDW and electrostatic,
>>>>>> significantly depend on rigid bonds are turned on or off. How it is
>>>>>> possible?
>>>>>>
>>>>>> Best regards,
>>>>>> --
>>>>>> Grzegorz Nawrocki
>>>>>>
>>>>>
>>>>> --
>>>>> Brian Radak
>>>>> Postdoctoral Scholar
>>>>> Gordon Center for Integrative Science, W323A
>>>>> Department of Biochemistry & Molecular Biology
>>>>> University of Chicago
>>>>> 929 E. 57th St.
>>>>> Chicago, IL 60637-1454
>>>>> Tel: 773/834-2812
>>>>> email: <radak_at_uchicago.edu>radak_at_uchicago.edu
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Grzegorz Nawrocki
>>>>
>>>
>>>
>>>
>>> --
>>> Brian Radak
>>> Postdoctoral Scholar
>>> University of Chicago
>>> Department of Biochemistry & Molecular Biology
>>> Gordon Center for Integrative Science, W323A
>>> 929 E. 57th St.
>>> Chicago, IL 60637-1454
>>> Tel: 773/834-2812
>>> e-mail: <radak_at_uchicago.edu>radak_at_uchicago.edu
>>>
>>
>>
>>
>> --
>> Grzegorz Nawrocki
>>
>>
>> --
>> Brian Radak
>> Postdoctoral Scholar
>> Gordon Center for Integrative Science, W323A
>> Department of Biochemistry & Molecular Biology
>> University of Chicago
>> 929 E. 57th St.
>> Chicago, IL 60637-1454
>> Tel: 773/834-2812
>> email: radak_at_uchicago.edu
>>
>
>
>
> --
> Grzegorz Nawrocki
>
>
> --
> Brian Radak
> Postdoctoral Scholar
> Gordon Center for Integrative Science, W323A
> Department of Biochemistry & Molecular Biology
> University of Chicago
> 929 E. 57th St.
> Chicago, IL 60637-1454
> Tel: 773/834-2812
> email: radak_at_uchicago.edu
>

-- 
Grzegorz Nawrocki

This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:21:26 CST