Re: I can't keep my pulling directions the same with the SMD vector

From: Axel Kohlmeyer (akohlmey_at_cmm.chem.upenn.edu)
Date: Tue Apr 21 2009 - 07:27:50 CDT

On Tue, 2009-04-21 at 10:49 +0300, Mert Gür wrote:
> Dear Axel thanks again for your insightfull answer. I really
> appreciate your help.
> But shouldn't I be able to define a PMF along any reaction coordinate
> I want? I mean considering different unbinding pathways I should be
> able to define a PMF for each case.

i don't know your system, so i cannot comment on this, but for
the sake of argument let us look at a simpler system. one that
i know very well (i've been working on these waaaaay back when
i was a graduate student): i want to compute the PMF (or rather
the solvent contribution) of adsorbing an ion to a metal surface.
if the metal surface is in x-y direction, the reaction coordinate
is the z-direction. now, all i have to look at is the z-component
of the forces on the ion, _regardless_ of how it diffuses in x-
or y-direction.

for your system, i'd first ask myself the question: is a straight
line a realistic reaction coordinate? or would a curved one be more
likely. if you pull fast with SMD, you can pull your system
along a straight line more easily, but would that be a correct
result? consider that you can usually get around a hill much
easier than across it, but if you have enough kinetic energy
it wouldn't matter.

perhaps giacomo or jerome can comment on whether the new collective
variable module in NAMD could be helpful in this case (using
meta-dynamic or ABF). there is now a large arsenal of free energy
methods available and each of them has its specific strengths and
weaknesses. rather than trying to force one method to work for your
problem, i'd rather evaluate which one would be the best match.

> >From what you are explaining I conclude that I either should perform
> multidimensional WHAM (which I have to figure how to do) or somehow
> define a reaction coordinate from which it wont deviate.

it can deviate, it just should not deviate systematically.

> Do you have any suggestion how to define that. Because that means that
> in my case I can' t perform US along the pulling direction I have
> performed for previously in my SMD simulations (Jarzynski).

see above for some suggestions, another thing that comes to my
mind would be the "string-method" which is effectively a kind
of self-adapting umbrella sampling.

cheers,
   axel.

> Best,
> Mert
>
>
> On Mon, Apr 20, 2009 at 4:41 PM, Axel Kohlmeyer
> <akohlmey_at_cmm.chem.upenn.edu> wrote:
> On Mon, 2009-04-20 at 11:16 +0300, Mert Gür wrote:
> > Sorry I didn't explain my problem properly.
> > As you emphasized I divided my reaction coordinates into
> bins. At each
> > specific coordinate (bin), I have to aplly an biasing
> potential to
> > keep it there. For that case I am using SMD and set the
> velocity to
> > zero. So I have an harmonic potential applied along my
> reaction
> > coordinate(x axis). But it does not stay on the SMD vector
> (1 0 0). As
> > you already mentioned it samples also into the other
> directions (y and
> > z directions) which as far as my understanding goes (by
> reading the
> > literature) we don't want. That is why I applied a negligble
> small
>
>
> mert,
>
> not in my understanding. if you do umbrella sampling you
> should
> not constrain any other direction but the reaction coordinate.
> if your system deviates from it randomly, it should average
> out,
> you just have to run long enough to make sure you are
> converged;
> if it deviates systematically, you have chosen the wrong
> reaction
> path. from all you are telling us, i am tempted to believe
> that
> the latter is the case.
>
> cheers,
> axel.
>
>
> > constraining potential into the x and y directions.
> > Using the positional fluctuation values along the reaction
> coordinates
> > at each US simulation I construct the PMF via WHAM.
> > Best,
> > Mert
> >
> >
> > On Mon, Apr 20, 2009 at 2:27 AM, Chris Harrison
> <char_at_ks.uiuc.edu>
> > wrote:
> > Mert,
> >
> > I'm not sure what you mean by "SMD with zero
> velocity for
> > Umbrella sampling.." In typical umbrella sampling
> the rxn
> > coordinate is divided into bins (10 bins is usually
> a
> > reasonable first guess if you have no a priori
> knowledge of
> > how many bins are needed) and each bin is sampled
> > independently; no SMD or pulling velocity needed.
> You then
> > "merge" the bins to construct to PMF via a weighted
> histogram
> > analysis. Not sure if this helped ....
> >
> >
> > C.
> >
> >
> >
> > --
> > Chris Harrison, Ph.D.
> > Theoretical and Computational Biophysics Group
> > NIH Resource for Macromolecular Modeling and
> Bioinformatics
> > Beckman Institute for Advanced Science and
> Technology
> > University of Illinois, 405 N. Mathews Ave., Urbana,
> IL 61801
> >
> > char_at_ks.uiuc.edu Voice:
> 217-244-1733
> > http://www.ks.uiuc.edu/~char Fax:
> 217-244-6078
> >
> >
> > Mert Gür <gurmert_at_gmail.com> writes:
> >
> > > Date: Sun, 19 Apr 2009 19:10:11 +0300
> > > From: Mert Gür <gurmert_at_gmail.com>
> >
> > > To: Chris Harrison <char_at_ks.uiuc.edu>
> > > Cc: NAMD list <namd-l_at_ks.uiuc.edu>
> > > Subject: Re: namd-l: I can't keep my pulling
> directions the
> > same with the SMD
> > > vector
> > > Return-Path: char_at_halifax.ks.uiuc.edu
> >
> > > Message-ID:
> >
> <72cbc58f0904190910g68238254kcae57f28652acb1_at_mail.gmail.com>
> > > X-Spam-Status: No, score=-1.1 required=5.0
> > tests=AWL,BAYES_00,
> > >
> DNS_FROM_RFC_POST,HTML_30_40,HTML_MESSAGE,SPF_PASS
> > autolearn=no
> > > version=3.1.7-0+tcb1
> > >
> >
> >
> > > Dear Chris,
> > > It seems that my x-values increases as I wish. As
> you
> > suggest pulling
> > > faster with a higher force constant seems to be a
> solution
> > (Now I have fixed
> > > the Jarzynski case :) ) but I also try to use SMD
> with zero
> > velocity for
> > > Umbrella Sampling Calculations.
> > > In that case the fast pulling option wont be
> possible.
> > > I should somehow to be able to sample the same
> reaction
> > coordinates and
> > > obtain a similar PMF for both methods.
> > > Best,
> > >
> > > On Sun, Apr 19, 2009 at 6:25 PM, Chris Harrison
> > <char_at_ks.uiuc.edu> wrote:
> > >
> > > > Mert,
> > > >
> > > > You need to look at the average x-coord value.
> Is it
> > increasing? There's
> > > > nothing that by default restrains the pulled
> particle to
> > specific values of
> > > > y & z when pulling in x. So while the x-coord
> value may
> > increase, the
> > > > particle may in fact freely sample the available
> y & z
> > coordinates. The
> > > > spring constant and force constant or pulling
> velocity can
> > notably influence
> > > > the degree or extent to which sampling can occur
> in y & z
> > for a given value
> > > > of x when pulling in x. Assuming a constant
> velocity SMD,
> > if you pull
> > > > faster with a strong spring constant the
> particle will
> > "not have enough time
> > > > and spring-flexibility" to sample y & z coords
> that
> > significantly deviate
> > > > from those of the initial structure ... but this
> will of
> > course lead to
> > > > poorer sampling and possibly a noisier force
> curve. It is
> > a balancing act,
> > > > the "tipping point" of which must be determined
> > empirically for each system.
> > > > One approach is to do some very fast pullings
> to test,
> > then lower your
> > > > pulling velocity and retest, then possibly
> refine your
> > spring constant to
> > > > achieve an optimal "signal-to-noise" in your
> force curve.
> > > >
> > > >
> > > > C.
> > > >
> > > >
> > > >
> > > > --
> > > > Chris Harrison, Ph.D.
> > > > Theoretical and Computational Biophysics Group
> > > > NIH Resource for Macromolecular Modeling and
> > Bioinformatics
> > > > Beckman Institute for Advanced Science and
> Technology
> > > > University of Illinois, 405 N. Mathews Ave.,
> Urbana, IL
> > 61801
> > > >
> > > > char_at_ks.uiuc.edu Voice:
> > 217-244-1733
> >
> > > > http://www.ks.uiuc.edu/~char
> <http://www.ks.uiuc.edu/%
> > 7Echar>
> >
> > > > Fax: 217-244-6078
> > > >
> > > >
> > > > Mert Gür <gurmert_at_gmail.com> writes:
> > > > > Date: Sun, 19 Apr 2009 15:56:00 +0300
> > > > > From: Mert Gür <gurmert_at_gmail.com>
> > > > > To: NAMD list <namd-l_at_ks.uiuc.edu>
> > > > > Subject: namd-l: I can't keep my pulling
> directions the
> > same with the SMD
> > > > > vector
> > > > > Return-Path: char_at_halifax.ks.uiuc.edu
> > > > > Message-ID:
> >
> <72cbc58f0904190556i3ed67eecke2b3c356b1b5c2e6_at_mail.gmail.com
> > > > >
> > > > > X-Spam-Status: No, score=-2.0 required=5.0
> > tests=AWL,BAYES_00,HTML_40_50,
> > > > > HTML_MESSAGE autolearn=unavailable
> version=3.1.7-0
> > +tcb1
> > > > >
> > > > > I try to evaluate th PMF of unbinding of a
> > protein-protein complex. As
> > > > was
> > > > > suggested for SMD calculations in the
> following paper, I
> > try to move only
> > > > > into the x direction. But I just can't keep my
> reaction
> > coordinates
> > > > (which
> > > > > is the vector between the fixed atoms and smd
> atoms)
> > along the x
> > > > direction.
> > > > > I also tried to aplly an external constraining
> force of
> > 1/10 in magnitude
> > > > of
> > > > > the SMD spring constant. But after each
> simulation the
> > vector keeps
> > > > > increasing its y magnitude.
> > > > > So my question here is how high can I go with
> these
> > constraining forces
> > > > so
> > > > > that my reaction coordinate moves strictly
> along the x
> > direction and
> > > > still
> > > > > makes sense. Or is there any other methodology
> I missed.
> > I tried to
> > > > follow
> > > > > all the discussion that have been performed
> previously
> > regarding the SMD
> > > > > calculations.
> > > > > Best,
> > > > > Mert
> > > > >
> > > > > Free energy calculation from steered molecular
> dynamics
> > simulations using
> > > > > Jarzynski's equality Park,
> > > > > Sanghyun<
> > > >
> http://adsabs.harvard.edu/cgi-bin/author_form?author=Park,
> > +S&fullauthor=Park,%
> 20Sanghyun&charset=UTF-8&db_key=PHY
> > > > >;
> > > > > Khalili-Araghi,
> > > > > Fatemeh<
> > > >
> >
> http://adsabs.harvard.edu/cgi-bin/author_form?author=Khalili-Araghi,+F&fullauthor=Khalili-Araghi,%20Fatemeh&charset=UTF-8&db_key=PHY
> > > > >;
> > > > > Tajkhorshid, Emad<
> > > >
> >
> http://adsabs.harvard.edu/cgi-bin/author_form?author=Tajkhorshid,+E&fullauthor=Tajkhorshid,%20Emad&charset=UTF-8&db_key=PHY
> > > > >;
> > > > > Schulten, Klaus<
> > > >
> >
> http://adsabs.harvard.edu/cgi-bin/author_form?author=Schulten,
> > +K&fullauthor=Schulten,%
> 20Klaus&charset=UTF-8&db_key=PHY
> > > > >
> > > > >
> > > > >
> > > > >
> http://adsabs.harvard.edu/abs/2003JChPh.119.3559P
> > > > >
> > > > > where it is indicated that:
> > > > >
> > > > > "In the simulation, we fix one end of the
> molecule ~the
> > N atom of the
> > > > first
> > > > > residue! at the origin and constrain the other
> end ~the
> > capping N atom at
> > > > > the C-terminus! to move only along the z axis,
> thereby
> > removing the
> > > > > irrelevant degrees of freedom, i.e., overall
> translation
> > and rotation"
> > > >
> >
> >
>
>
> --
> =======================================================================
> Axel Kohlmeyer akohlmey_at_cmm.chem.upenn.edu
> http://www.cmm.upenn.edu
> Center for Molecular Modeling -- University of
> Pennsylvania
> Department of Chemistry, 231 S.34th Street, Philadelphia, PA
> 19104-6323
> tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel:
> 1-215-898-5425
> =======================================================================
> If you make something idiot-proof, the universe creates a
> better idiot.
>
>

-- 
=======================================================================
Axel Kohlmeyer   akohlmey_at_cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:50:48 CST