[gmx-users] ligand moving out during umbrella sampling

[Justin Lemkul]

On 5/8/17 10:00 AM, abhisek Mondal wrote:
> On Sun, May 7, 2017 at 11:37 PM, Justin Lemkul <jalemkul at vt.edu> wrote:
>
>>
>>
>> On 5/7/17 1:57 AM, abhisek Mondal wrote:
>>
>>> Hi,
>>>
>>> For your ease of understanding regarding what is happening during this
>>> above said umbrella-mdrun, I have shared the trajectory video file the
>>> following link.
>>> https://drive.google.com/drive/folders/0B6O-L5Y7BiGJQ1FIc2tIRFE2dE0
>>>
>>> Is this normal given that the mdp code being used ? I basically have no
>>> idea with this step, so please help me out. I'm using gromacs-4.6.2.
>>>
>>>
>> Your setup is incorrect.  You're applying a biasing potential only along
>> z, so the ligand can move freely along x and y.  A protein-ligand complex
>> has spherical symmetry, so you should set the reaction coordinate to the
>> vector connecting the ligand with some suitable subset of interacting
>> protein residues.
>
>
> I don't get it.
> We are trying not to move our configuration (generated after pulling
> simulation) along the reaction coordinate, so for restraining we are
> supposed to set pull_rate1=0.0.

Of course.  But you said set pull_k = 0 which does not make sense.  The pulling 
rate *is* zero during umbrella sampling (no net displacement, restrain to the 
specified distance along the reaction coordinate) and pulling force constant 
should be non-zero.

> If applying biasing potential only along z is causing movement along x and
> y then what if we apply the biasing potential along x,y,z ? Will it cause
> any good in restraining the ligand?
>

This is how it should be done.  The reaction coordinate should be suitably 
defined based on the geometry of the system.  As I suggested before, choose some 
representative residues in the active site as one group and the ligand as the 
other.  Thus defines the reaction coordinate without any presupposition of 
anything being aligned with a Cartesian axis, which is rarely the case.

> Moreover, you said previously "A protein-ligand complex has spherical
> symmetry, so you should set the reaction coordinate to the vector
> connecting the ligand with some suitable subset of interacting protein
> residues.". It is really unclear to me, could you please give me some
> examples to understand it more simply? I had pulled the ligand along -z
> axis, doesn't it mean that the reaction coordinate is to be that way ? The
> fact that I'm struggling with is to restrain the pull configurations for
> further sampling.
>

The reaction coordinate is whatever you define it to be.  Whether or not pulling 
along the z-axis makes sense depends on the orientation of the system and the 
intrinsic geometry.  In your case, it doesn't make sense.  In my case (the 
tutorial, the unidirectional growth of an amyloid fibril) it does make sense to 
use a single Cartesian axis for the SMD portion and subsequent umbrella sampling.

> I'm really a beginner, so maybe I'm asking stupid questions. Please give me
> some advise. I'm really unable to decipher the scenario in comparison to
> your amyloid article in JPCB.
>

You should read the article to understand why I did what I did in the tutorial, 
and then move on to reading articles that are more similar to your case.  These 
will be much more relevant to what you're doing.

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================