Re: Proposed Feature in Targeted MD

From: felmerino_at_uchile.cl
Date: Mon Jan 31 2011 - 20:20:04 CST

Hey,

I tried the method that jerome proposed with metadynamics and it appears to work really well (they are preliminary tests although). When i did the same using a SMD style simulation with that same colvar to calculate work the results were very poor. Also, there is a paper somehow benchmarking metadynamics and abf in namd where a similar approach is performed with abf (if i remember well it is a henin et alii 2010.... i'm not sure).

If you are interested on the pmf along a certian path defined by some end points maybe you should try the path collective variables available in plumed. You can mix that code with your favourite md engine (including NAMD and Gromacs)

best

felipe

----Mensaje original----
De: jhenin_at_ifr88.cnrs-mrs.fr
Fecha: 31-ene-2011 19:39
Para: "V.Ovchinnikov"<ovchinnv_at_mit.edu>
CC: "Minh, David D."<daveminh_at_gmail.com>, <namd-l_at_ks.uiuc.edu>
Asunto: Re: namd-l: Proposed Feature in Targeted MD

Hi guys,

Try the colvars (collective variables) module of NAMD 2.7: you can
define a colvar as a difference of two RMSD values. Then you can apply
a moving restraint to that variable, or even run metadynamics.

Cheers,
Jerome

On 31 January 2011 22:53, V.Ovchinnikov <ovchinnv_at_mit.edu> wrote:
> You are right about the compilation issue. However, I also have a
> version of TMD that I used in a TCL script (for which you obviously need
> just the binary); this was a somewhat painful exercise that I undertook
> in 2006 because the official NAMD 2.6 version of TMD had a bug which was
> visible only in parallel runs, (apparently, I wasn't clever enough to
> fix the source code &amp; recompile). The bug was fixed a couple of years
> ago.
>
> ( If you are wondering, the consuming part in tcl was coding the
> best-fit rotation a la Kabsch 76, which involves diagonalization )
>
> So, while you figure out how to compile the source, I can also send to
> you the TCL script (it is not slow if the number of forced atoms is
> small, e.g. if you run an explicit water simulation and apply the
> restraint to the protein atoms only)
>
> Will be in touch,
> Victor
>
> On Mon, 2011-01-31 at 15:33 -0600, Minh, David D. wrote:
>> Dear Victor,
>>
>> Thanks for the fast reply!
>>
>> Yes, it would be quite helpful to see your code. Please send it when you find the time. If there is interest in the community, I could help polish this code for later inclusion.
>>
>> I still need a way to compile the source code on my Mac, though. I've only worked with the binaries before. Hopefully somebody else knows what I can do about that problem.
>>
>> Thanks!
>> David
>>
>> On Jan 31, 2011, at 3:14 PM, V.Ovchinnikov wrote:
>>
>> > Dear David,
>> >
>> > I have tinkered with the restrained version of TMD that is in NAMD2.6 a
>> > couple of years ago, and I think I have a version that performs a
>> > double-sided TMD using two orientation structures, as you describe
>> > (although I was not using it to compute work, just to look possible
>> > paths). If you give me a couple of days to dig up the relevant files, I
>> > can send them to you off-the-list. Not quite sure how the TMD 2.7 source
>> > code is different from TMD 2.6, but I would be surprised if the
>> > differences were large. (One could polish the code and submit to the
>> > developers for inclusion later, too, if the community thinks its
>> > useful.)
>> >
>> > Regards,
>> > Victor
>> >
>> >
>> > On Mon, 2011-01-31 at 13:12 -0600, Minh, David D. wrote:
>> >> A minor change to targeted MD can make it possible to calculate
>> >> potentials of mean force in the region between two structures.
>> >>
>> >>
>> >> Right now, the code allows targeting toward one final structure, with
>> >> the potential given by the following equation,
>> >> U_{TMD} = \frac{1}{2} \frac{k}{N} [ RMS(t) - RMS^*(t) ]^2
>> >> http://www.ks.uiuc.edu/Research/namd/2.7/ug/node41.html
>> >>
>> >>
>> >> It should be relatively easy to modify targeted MD and have the option
>> >> of targeting the transition between two structures. The potential
>> >> would be,
>> >> U_{TMD} = \frac{1}{2} \frac{k}{N} [ DRMS(t) - DRMS^*(t) ]^2,
>> >> with DRMS(t) = RMS_1(t) - RMS_2(t), and RMS_1 being the RMSD from
>> >> structure 1 and RMS_2 being the RMSD from structure 2, and DRMS^*(t)
>> >> being a time-dependent function. This bias was used in,
>> >> NK Banavali and B Roux. Free Energy Landscape of A-DNA to B-DNA
>> >> Conversion in Aqueous Solution. J. Am. Chem. Soc. 127(18), 6966-6876
>> >> (2005).
>> >> K Arora and CL Brooks. Large-scale allosteric conformational
>> >> transitions of adenylate kinase appear to involve a population-shift
>> >> mechanism. PNAS 104(47), 18496-18501 (2007).
>> >> and probably other papers.
>> >>
>> >>
>> >> Using the theory described in,
>> >> G Hummer and A Szabo. Free energy reconstruction from nonequilibrium
>> >> single-molecule pulling experiments. PNAS 98(7), 3658-61 (2001).
>> >> one could calculate the PMF along the DRMS coordinate from a set of
>> >> TMD trajectories. The theory described in,
>> >> DDL Minh. Multidimensional Potentials of Mean Force from Biased
>> >> Experiments along a Single Coordinate. J. Phys. Chem. B 111,
>> >> 4137-4140 (2007).
>> >> can be used to calculate the PMF along other coordinates, or 2D PMFs.
>> >> The theory in,
>> >> DDL Minh and AB Adib. Optimized Free Energies from Bidirectional
>> >> Single-Molecule Force Spectroscopy. Phys. Rev. Lett 100(18), 180602
>> >> (2008).
>> >> DDL Minh and JD Chodera. Optimal estimators and asymptotic variances
>> >> for nonequilibrium path-ensemble averages. J. Chem. Phys. 131(13),
>> >> 134110 (2009).
>> >> would be useful for combining data for the transition from structure 1
>> >> to structure 2 and the reverse process from structure 2 to structure
>> >> 1.
>> >>
>> >>
>> >> Actually, it is possible to reconstruct PMFs with the current TMD
>> >> code, but I think the PMF between two structures would be more
>> >> interesting.
>> >>
>> >>
>> >> I'd like to give this a try. Would any NAMD developers be interested
>> >> in making this modification? Any thoughts?
>> >
>>
>
>

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:19:46 CST