[gmx-users] Umbrella sampling and PMF calculations

[Justin Lemkul]

On 6/2/17 4:17 PM, Alex wrote:
> So, Justin, just to follow up here. With 'cylinder', grompp refuses to
> accept a zero pulling rate. This begs a different question: in my

I find that surprising.  Please provide the full, exact error and your pull 
settings.  All of the geometries should work with zero or non-zero pull rates. 
But I admit, it has been many years since I used GROMACS with the pull code.

> particular system, the COM position coincides with the center of the single
> reactive pore at the geometric center of the membrane. Why even bother with
> this setup? I'd understand if the entire membrane was reactive, but it
> isn't. Also, the obtained free energy value is very close to what one gets
> by simply integrating the pullf data wrt pull coordinate, which makes
> perfect sense. Any comments/suggestions?
> 

You might get some windows that conveniently work with "simpler" settings but I 
still think you need to do what I said before.  Old GROMACS versions had a 
"position" geometry that was the go-to for membrane/slab/layered systems and 
worked quite nicely, but that's gone now, replaced by other options that I've 
never used.

-Justin

> Thanks,
> 
> Alex
> 
> On Fri, Jun 2, 2017 at 11:26 AM, Alex <nedomacho at gmail.com> wrote:
> 
>>
>>
>>>>
>>> Yes, because if your ion diffuses laterally (or whatever constitutes your
>>> "hole"), then you're not sampling what you think you're sampling.  That's
>>> what the cylinder geometry is for.  You have a system capable of
>>> significant lateral movement, so if you fail to apply a bias that acts in
>>> the x-y plane, your results are very quickly going to become garbage.
>>>
>>> -Justin
>>>
>>> Of course. I actually realized that you used COM distance and not e.g.
>> direction-periodic (which won't accept zero pulling rates anyway) only
>> after I asked.
>> Specifying somewhat awkward structures like disks as references could be
>> avoided by simply allowing a component of the COM distance vector and not
>> the entire vector) to be used. Anyway, doesn't really matter.
>>
>> Thanks,
>>
>> Alex
>>

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

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

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