Hi,
It is known that reweighting aMD simulations using the boost energies
becomes very difficult as the applied boost increases, see e.g.:
Markwick and McCammon, Studying functional dynamics in bio-molecules
using accelerated molecular dynamics, Phys. Chem. Chem. Phys.,
2011,13, 20053-20065.
The problem basically boils down what they refer to as 'statistical
noise error' which results from the boost potential, and 'statistical
mechanical sampling error', which essentially means you need
relatively complete sampling of the modified potential surface - these
errors are unfortunately in opposition to one another.
I think a good way to address this issue is to combine aMD with
H-REMD, where you have one unbiased Hamiltonian and the rest have aMD
with increasing levels of boost. This approach has been used for a
while for speeding up convergence in free energy calculations, and we
have recently used this approach in studying convergence of an RNA
tetranucleotide system:
http://pubs.acs.org/doi/abs/10.1021/jp4125099
Hope this helps,
-Dan
On Sun, Apr 27, 2014 at 5:26 PM, Thomas Evangelidis <tevang3.gmail.com> wrote:
> Hi Vlad,
>
> How did you decide that the trajectories could not be reweighted?
> Στις 28 Απρ 2014 12:53 ΠΜ, ο χρήστης "Vlad Cojocaru" <
> vlad.cojocaru.mpi-muenster.mpg.de> έγραψε:
>
>> Hi Jef, Hi Sourav,
>>
>> Maybe I can comment further here ...
>>
>> We did quite a bunch of aMD (only dihedral boost) runs on
>> protein/nucleic acid complexes (not published yet) and I can confirm
>> that it works well at least qualitatively. Using parameters calculated
>> based on the suggestions on the amber manual (3.5 kcal/mol/residue to
>> estimate alpha and Ed) did not affect DNA much. However, adding 2 alphas
>> to the boost did destroy the DNA at the ends on some of our systems.
>>
>> However, the problem came when we tried to reweight. We realized that
>> the boost was far too high for reweighting even using the second order
>> cumulant expansion.
>> We came across this paper that I mentioned in my previous email ...
>> http://www.ncbi.nlm.nih.gov/pubmed/23781107 and we switched to apply the
>> lambda-based scheme for estimating the parameters. The paper suggests
>> lambda=0.3 for the protein/membrane system they investigated. However,
>> we found 0.3 to be too large or our systems as the reweigthing was still
>> compromised. We now switched to test different lamdas and a value of
>> 0.15 appears to give very nice trajectories that sample significantly
>> more the dynamics of the systems compared with standard MD and can also
>> be reweighted. However, we also noticed that in order to get smooth
>> profiles upon reweighting we need to run quite long simulations (150 ns
>> is by far not enough as the reweighted profiles are still noisy).
>>
>> This is to share our experience with the particular systems we work
>> with. Whether this is applicable to other system, I don't know.
>> So,please test for any different system. But I would definitely
>> recommend the lambda-based scheme to estimate the parameters (its only 1
>> variable to be changed, and therefore easier to run tests with)...
>>
>> I hope this helps
>> Vlad
>>
>> On 04/27/2014 09:16 PM, Jeff Wereszczynski wrote:
>> > Hi Sourav,
>> >
>> > I haven't looked at the suggestions in the Amber user guide, but I have a
>> > pretty good guess as to what they are. It sounds like you are doing a
>> > "dual-boost" setup. So here's what I would say.
>> >
>> > 1. Yes. The number of atoms is typically used for the term that boost
>> the
>> > complete potential, which is dominated by non bonded terms. Since waters
>> > are included in that, you need to include the water atoms.
>> >
>> > 2. As far as I know, no one has published aMD simulations with nucleic
>> > acids. I've played with it a bit, and it works fine. You should include
>> > the DNA in the residue count since they are being accelerated by the
>> > dihedral term (since they have dihedrals in them). Just keep in mind
>> that
>> > the suggested aMD parameters were derived mainly for proteins with the
>> > amber force field, so you may need to fiddle with the acceleration
>> > parameters you use. But the suggestions in the amber user guide should
>> get
>> > you in the right ballpark.
>> >
>> > Typically what you want to do is try a number of short simulations with
>> > different parameters, using the parameters in the amber user guide as a
>> > suggestion, then seeing how it affects your system. If you need more
>> > acceleration, try adding a value of alpha (or alphaD) to your thresholds.
>> > If you have too much acceleration and you are getting unphysical
>> effects,
>> > decrease your thresholds by alpha. Other people have different
>> > parameterization methods, but in the past this has worked well for me.
>> >
>> > Cheers,
>> >
>> > Jeff Wereszczynski
>> > Assistant Professor of Physics
>> > Illinois Institute of Technology
>> > http://www.iit.edu/~jwereszc
>> >
>> >
>> >
>> >> Message: 2
>> >> Date: Sun, 27 Apr 2014 16:31:08 +0530
>> >> From: Sourav Purohit <sour000.gmail.com>
>> >> Subject: [AMBER] Accelerated MD
>> >> To: AMBER Mailing List <amber.ambermd.org>, Jason Swails
>> >> <jason.swails.gmail.com>, ross.rosswalker.co.uk
>> >> Message-ID:
>> >> <CAFMrzpzPjf1ZN2-V71v=fkeZrDeGNytZsAQSpL7ZxM1xysvs=
>> >> w.mail.gmail.com>
>> >> Content-Type: text/plain; charset=ISO-8859-1
>> >>
>> >> Hi all,
>> >>
>> >> I had some doubts on the method of calculating parameters for
>> accelerated
>> >> MD as described in the AMBER12 manual (page no.160).
>> >>
>> >> 1.Does the total number of atoms (taken as 16950 in the example) include
>> >> atoms belonging to water molecules also??
>> >>
>> >> 2.for the calculation of alphaD, only the number of protein residues
>> goes
>> >> as an input. What if a system has other components like DNA?? Should I
>> >> include the DNA residues also??
>> >>
>> >> Kindly help.
>> >>
>> >> Thanks.
>> >>
>> >>
>> >>
>> > _______________________________________________
>> > AMBER mailing list
>> > AMBER.ambermd.org
>> > http://lists.ambermd.org/mailman/listinfo/amber
>> >
>> >
>>
>> --
>> Dr. Vlad Cojocaru
>> Max Planck Institute for Molecular Biomedicine
>> Department of Cell and Developmental Biology
>> Röntgenstrasse 20, 48149 Münster, Germany
>> Tel: +49-251-70365-324; Fax: +49-251-70365-399
>> Email: vlad.cojocaru[at]mpi-muenster.mpg.de
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
--
-------------------------
Daniel R. Roe, PhD
Department of Medicinal Chemistry
University of Utah
30 South 2000 East, Room 201
Salt Lake City, UT 84112-5820
http://home.chpc.utah.edu/~cheatham/
(801) 587-9652
(801) 585-6208 (Fax)
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Apr 27 2014 - 17:30:02 PDT
