VMD-L Mailing List
From: Mayne, Christopher G (cmayne2_at_illinois.edu)
Date: Fri Mar 20 2015 - 13:19:51 CDT
The values are not fixed. They are updated at each iteration of the optimization. What the user is requesting is NOT e.g.:
k(1) = 40.0, b0(1) = 1.2
k(2) = 40.0, b0(1) = 1.2
where 40.0 and 1.2 are specified by the user.
but rather IS requesting this,
k(1) = k(2) = <some force constant passed in by the optimizer>
b0(1) = b0(2) = <some distance passed in by the optimizer>
The former is more appropriately handled by assigning the parameters in an "Associated Parameter File" and excluding the terms from the optimization. The latter requires significant and non-trivial modification of the existing optimizer functions.
On Mar 20, 2015, at 1:06 PM, Richard Wood wrote:
> But fixing them to the same value makes them equivalent, no? Or does 6 not equal 6?
> From: Mayne, Christopher G [cmayne2_at_illinois.edu]
> Sent: Friday, March 20, 2015 12:53 PM
> To: Richard Wood
> Cc: VMD Mailing List
> Subject: Re: vmd-l: fftk bonded and dihedral optimization question
> The point of ffTK is optimize those parameters; they are not known a priori. The constraint is not to fix the parameters to a specific value, but rather to make them equivalent, whatever that value may be.
> On Mar 20, 2015, at 12:50 PM, Richard Wood wrote:
>> Her original post said the following: "...For example, if I have two bonds, CD2D1 - CD2D1 and CD2D2-CD2D2 and I want them both to have the same force constant and equilibrium distance. "
>> I would give both bonds the same parameters.
>> From: Mayne, Christopher G [cmayne2_at_illinois.edu]
>> Sent: Friday, March 20, 2015 12:41 PM
>> To: Richard Wood
>> Cc: VMD Mailing List
>> Subject: Re: vmd-l: fftk bonded and dihedral optimization question
>> By my understanding the request is not to fix the bond to a specific value, but rather to add an additional constraint to the optimization that assigns equivalent parameters to these particular bond terms. This is a different issue that requires special handling by the optimizer.
>> Christopher Mayne
>> On Mar 20, 2015, at 12:27 PM, Richard Wood wrote:
>>> But all she wants to do is constrain the two bonds to a constant length, which I have done many times. I did it by setting R0 in the parameter file to the value I wanted and I also made the force constant very large.
>>> From: owner-vmd-l_at_ks.uiuc.edu [owner-vmd-l_at_ks.uiuc.edu] on behalf of Mayne, Christopher G [cmayne2_at_illinois.edu]
>>> Sent: Friday, March 20, 2015 11:21 AM
>>> To: Tim Lo
>>> Cc: Chaya Stern; <vmd-l_at_ks.uiuc.edu>
>>> Subject: Re: vmd-l: fftk bonded and dihedral optimization question
>>> Tim and Chaya,
>>> There is not currently a way to assign these types of constraints, although this is something that is on our TODO list. I would urge caution with regard to manually assigning atom types to the same type as a mechanism of circumventing this problem. There is generally a reason why the atom types were assigned differently (e.g., LJ terms).
>>> Christopher Mayne
>>> On Mar 20, 2015, at 10:36 AM, Tim Lo wrote:
>>>> If all the parameters related to CD2D1 and CD2D2 are equal, you can simply make them to be the same atom type. If this is not the case, do you mean that most parameters (e.g. LJ parameters) related to CD2D1 are different from those related to CD2D2 while only some parameters (e.g. CD2D1 - CD2D1 and CD2D2-CD2D2) are the same? Then, be careful as there may be logical loopholes. There must be some physical reasons (e.g. experimental evidence) to allow you to manually assign most parameters to be the same while some parameters are different. Different atom types usually have different parameters (except for nonbonded parameters). Manually assigning (or in your word, restraining) parameters is not a good way to do parametrization (except in some cases where nonbonded parameters are assigned by analogy).
>>>> Research Assistant
>>>> The University of Hong Kong
>>>> On 2015/3/20 $B2<8a(B 10:13, Chaya Stern wrote:
>>>>> Is there any way to restrain two bonds or angles or dihedrals to be equivalent when I optimize the parameters? For example, if I have two bonds, CD2D1 - CD2D1 and CD2D2-CD2D2 and I want them both to have the same force constant and equilibrium distance.
>>>>> Thank you,
>>>>> Chaya D Stern
>>>>> Memorial Sloan Kettering Cancer Center
>>>>> Tri-Institutional Chemical Biology Program