# Re: Energy minimization in NAMD

From: Ajasja Ljubetič (ajasja.ljubetic_at_gmail.com)
Date: Tue Aug 09 2016 - 11:27:41 CDT

This depends on the starting conformation of your molecule. Local
minimization in general does not go "up-hill".
Even if you have just one dihedral angle you are not guaranteed to reach
the global minimum. For example if you start local minimization from the
red point in the image you will always reach the left local minimum, never
the global. Although finding the global minimum is rarely required for MD
simulations (given enough sampling time).

[image: Inline images 1]
This are very basic questions, so hopefully you have a local advisor who
can help you. Reading a general book on simulations would also not hurt.

FYI, you can also do simulated annealing in NAMD.

Best,
Ajasja

On 9 August 2016 at 18:09, Faramarz Joodaki <fjoodaki_at_my.uri.edu> wrote:

> Dear Ajasja,
>
> If we have a just one molecule with 27 atoms, can we reach to global
> minimum by NAMD minimization or we need to apply other methods?
>
> Best Regards,
> Faramarz
>
> On Tue, Aug 9, 2016 at 11:01 AM, Faramarz Joodaki <fjoodaki_at_my.uri.edu>
> wrote:
>
>> Dear Ajasja,
>>
>> Thank you for your explanation!
>>
>> Best Regards,
>> Faramarz
>>
>> On Tue, Aug 9, 2016 at 5:24 AM, Ajasja Ljubetič <
>> ajasja.ljubetic_at_gmail.com> wrote:
>>
>>> Sure,
>>>
>>> Your system is stuck in a local minimum/saddle. Local minimization
>>> methods (such as conjugated gradient etc...) will not help you get out of
>>> the local minimum. You can try to heat up the system and cool it down
>>> again. This is called simulated annealing
>>> cross from the local to the global minimum.
>>>
>>> Alternately you can very the dihedral angles of your molecule, so that
>>> you start local minimization from different conformations, one of which
>>> will hopefully converge to the global minimum.
>>> Although rigorously proving that you have found the global minimum is
>>> also difficult.
>>>
>>> Good luck,
>>> Ajasja
>>>
>>>
>>> On 8 August 2016 at 23:42, Faramarz Joodaki <fjoodaki_at_my.uri.edu> wrote:
>>>
>>>> Dear Ajasja,
>>>>
>>>> Thank you for your email!
>>>> Unfortunately, I did not get your idea! Could you please explain a
>>>> little more?
>>>>
>>>> Best Regards,
>>>> Faramarz
>>>>
>>>> On Mon, Aug 8, 2016 at 4:54 PM, Ajasja Ljubetič <
>>>> ajasja.ljubetic_at_gmail.com> wrote:
>>>>
>>>>> Hi!
>>>>>
>>>>> You could try simulated annealing
>>>>> <https://www-s.ks.uiuc.edu/Research/namd/mailing_list/namd-l.2003-2004/0034.html>
>>>>> or simply generating large amounts of starting conformations far apart and
>>>>> minimizing them all.
>>>>>
>>>>> Best,
>>>>> Ajasja
>>>>>
>>>>> On 8 August 2016 at 18:37, Faramarz Joodaki <fjoodaki_at_my.uri.edu>
>>>>> wrote:
>>>>>
>>>>>> Hi everyone!
>>>>>>
>>>>>> I have done 10000 steps energy minimization on my system (1599
>>>>>> atoms). According to total energy plot and the amount of energy, I can
>>>>>> state that the energy is stable. But I calculated the eigenvalues of the
>>>>>> system (after minimization) and I found that some negative eigenvalues
>>>>>> which it means that I am in the saddle point instead of global minimum. By
>>>>>> changing the minimization parameters I could not reach the global minimum
>>>>>> point. It is really important for me to reach that point for my next step
>>>>>> research.
>>>>>>
>>>>>> I found also some parameters in the log file, which I do not know
>>>>>> exactly what they mean:
>>>>>>
>>>>>> *DX, DU and DUDX.*
>>>>>> In addition, I found that NAMD has just two methods for minimization--001a114b7d323216580539a60343--
--001a114b7d3232165a0539a60344--

This archive was generated by hypermail 2.1.6 : Sun Dec 31 2017 - 23:20:37 CST