[gmx-users] Re: distance/direction PMF

[Gavin Melaugh]

Hi Justin

I think what is happening is that grompp is returning the projection of
the current vector between the two reference groups onto the stated
vector in the mdp file. i.e the dot product of the two vectors. The
resulting PMF plot must therefore be  a function of this varying cross
product.  I say varying as I don't think the dynamics constrain the pull
group alone that vector alone.
When you change the vector in the mdp file to be that of the new vector
connecting the two groups grompp gives the actual distance between the
two groups.

Gavin

Justin A. Lemkul wrote:
>
>
> Gavin Melaugh wrote:
>> Yeah
>>
>>  I set pull_init =0, and pull_start = yes, and it still gave a "distance
>> at start" as 0.78nm. It must be doing something funny because with
>> pull_vec, because when I use pull_geometry = distance and pull_dim
>> = Y Y Y, the distance 0.815nm is returned as the "distance at start",
>> which is the actual distance between the two groups.
>>
>
> I have no idea why this is happening.  If you'd like me to
> troubleshoot further and see if I can get to the bottom of this,
> please send me (off-list is OK):
>
> 1. Coordinate file
> 2. Topology file(s)
> 3. The .mdp file that is giving the weird result
>
> -Justin
>
>> Cheers
>>
>> Gavin
>>  
>> Justin A. Lemkul wrote:
>>>
>>> Gavin Melaugh wrote:
>>>> Hi Justin
>>>>
>>>> Sorry maybe I was unclear before with my first point; the specified
>>>> distance in the mdp file is 0.78nm, which grompp returns back as
>>>> the ref
>>>> distance. The "distance at the start" that grompp returns is also
>>>> 0.78nm, even though the actual distance is 0.815nm. That's why I was
>>>> asking what this distance actually means, as it is not the absolute
>>>> distance between the reference groups.
>>> As I said before, you're telling grompp to make that the reference
>>> distance. You say "the specified distance in the mdp file is 0.78 nm,"
>>> therefore grompp is doing what it's told.  If that's not the desired
>>> distance, then don't set it as such.  Have you tried the combination I
>>> suggested in the last message?  Does it give an initial distance of
>>> 0.815 nm?
>>>
>>>> Also as regards to your second point; what if in the initial
>>>> configurations the two reference groups do not lie along the stated
>>>> vector?
>>>>
>>> Hard to say, but probably you'll get some very high forces at the
>>> outset of the simulation as the simulation as the umbrella potential
>>> tries to force the pulled group to conform to the restraint
>>> specifications.  Whether or not that negatively impacts the ultimate
>>> result of the simulation is also an important consideration.  If
>>> you're pulling along a given vector, you should start with your
>>> reference coordinates as close to the desired location as possible.
>>>
>>> -Justin
>>>
>>>> Cheers
>>>>
>>>> Gavin
>>>>
>>>> Justin A. Lemkul wrote:
>>>>> Gavin Melaugh wrote:
>>>>>> Justin
>>>>>>
>>>>>> I did a short test on a particular window with the specified vector
>>>>>> with
>>>>>> pull_init =0.78nm. I then grompp(ed) the final configuration and it
>>>>>> gave
>>>>>> me a distance of 0.78 as the initial ditsance, fair enough. However
>>>>>> when
>>>>>> I viewed the final line from g_dist the absolute distance or
>>>>>> modulus was
>>>>>> 0.815 nm, which agreed with the distance I calculated using a
>>>>>> molecular
>>>>>> configuration editor. My question is therefore this, when using
>>>>>> pull_geometry =direction with a specified vector, How should one
>>>>>> interpret the initial distance provided by grompp? and how does
>>>>>> Gromacs
>>>>>> deal with the fact that the vector between the two point changes
>>>>>> during
>>>>>> the run ?
>>>>>>
>>>>> You're telling grompp that the initial distance is 0.78 nm, so it's
>>>>> spitting that back out.  If that's not the true distance, then
>>>>> specify
>>>>> the correct quantity :)  The distance should be correctly detected
>>>>> with:
>>>>>
>>>>> pull_start = yes
>>>>> pull_init = 0
>>>>>
>>>>> As for the second question, the vector doesn't change over the
>>>>> simulation, but the position of pull_group1 along that vector will.
>>>>> That's what the umbrella potential is doing - it allows for harmonic
>>>>> oscillation along any of the dimensions specified by
>>>>> pull_vec/pull_dim
>>>>> (depending on the settings).
>>>>>
>>>>> -Justin
>>>>>
>>
>