From: Hadi (dinpajooh_at_gmail.com)
Date: Tue Aug 05 2014 - 14:49:48 CDT
On Tue, Aug 5, 2014 at 12:29 PM, Kenno Vanommeslaeghe <
> Two questions:
> (1) knowing that computers don't have infinite precision, there will
> *always* be small deviations. The question to ask is: are they significant?
> (2) Did you plot the same distribution for an MC simulation?
Yes. My MC code does produce a distribution which is a delta function. As I
mentioned, the NAMD results have slight deviations from a delta function.
However, I quite agree with you: in addition to the previous factors
(algorithm, rigid tolerance, time step), precision should be considered as
> Because I see no fundamental reason the distribution of rigid bond lengths
> in MD should have a higher sigma than MC.
Sure, one can add rigid body rotations to an MC move set, but one could
> also implement rigid bodies in MD (albeit at the cost of increased
> complexity). Conversely, one could conceivably use something similar to
> SHAKE in MC. SHAKE essentially just calculates the rigid body translation
> and rotation numerically (as opposed to analytically).
> On 08/05/2014 03:10 PM, Hadi wrote:
>> In general, I think it is fair to say that unlike MC simulations, the
>> rigid models in MD simulations are only slightly flexible considering the
>> algorithm, rigid tolerance, time step, etc. that will be used. I have
>> plotted the histogram of bond length for a given model with appropriate
>> parameters and observed that the distribution of bond length has some
>> small deviations from a delta function.
>> On Tue, Aug 5, 2014 at 11:41 AM, Kenno Vanommeslaeghe
>> <kvanomme_at_rx.umaryland.edu <mailto:kvanomme_at_rx.umaryland.edu>> wrote:
>> That doesn't mean the model wouldn't be rigid, just that the
>> temperature wouldn't be right. More importantly, I don't think the
>> person who asked the original question is planning to perform Langevin
>> integration. Langevin is good to mimic the effects of solvent
>> collisions when running implicit solvent and can also be used to
>> improve sampling in some instances, bit it wouldn't be very
>> appropriate for his purpose. The gold standard for explicit solvent
>> simulations with CHARMM is Verlet (and that's also what was used to
>> parametrize the force field).
>> On 08/05/2014 12:52 PM, Hadi Dinpajooh wrote:
>> Well, as mentioned in the link NAMD is not well developed if one
>> rigid models and Langevin integration:
>> " ... Although /NAMD/ will run with Langevin integration and rigid
>> both active, the equilibrium temperature of the system does not
>> to the correct value. Different Langevin equations are needed to
>> the correct velocity distribution, which have not yet been added
>> to /NAMD/."
>> Sent from my iPhone
>> On Aug 5, 2014, at 7:54 AM, Kenno Vanommeslaeghe
>> <kvanomme_at_rx.umaryland.edu <mailto:kvanomme_at_rx.umaryland.edu>
>> <mailto:kvanomme_at_rx.umaryland.edu>>> wrote:
>> On 08/05/2014 03:14 AM, Hadi wrote:
>> (2) It is my understanding that the parameters you
>> copy-pasted into
>> your e-mail are just "reasonable values" for in case
>> an MD engine or
>> methodology is used that does not support rigid water.
>> They won't give
>> the exact same properties as the true rigid TIP3P
>> model. NAMD,
>> however, does support rigid water; assuming the psf is
>> correctly, all one need to do is put the following
>> line in the NAMD
>> rigidBonds water
>> Though if you're using the CHARMM force field, it's
>> recommended to instead use
>> rigidBonds all
>> because that's what the force field is parametrized
>> I hope it is clear that unlike MC simulations, a rigid
>> model can actually
>> be slightly flexible in MD simulations.
>> Umm no, provided that no mistakes were made in the set-up of
>> (rigidBonds in NAMD), a rigid model is really rigid in MD.
>> Except for
>> insignificant numerical deviations related to the SHAKE
>> tolerance. More
>> info here:
This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:42 CST