Re: Translocation problem while attempting to use ABF

From: Jerome Henin (jhenin_at_cmm.chem.upenn.edu)
Date: Mon Jul 06 2009 - 11:37:20 CDT

Hi Hugh,

You could be running into a problem with the way the abf code form
NAMD 2.6 handles PBC - that is, it mostly doesn't. One thing you can
look at is the role of the boundary potentials ("walls") on your
problem. Try setting forceConst to zero and see if the problem still
occurs. That's all I can say based on the info you provided.

Note that this code is currently being phased out, and replaced with
the version of ABF implemented in NAMD 2.7b1's module called
collective variable-based calculations:
http://www.ks.uiuc.edu/Research/namd/2.7b1/ug/node46.html

Best,
Jerome

On Mon, Jul 6, 2009 at 11:39 AM, Hugh Martin<hughtendo_at_gmail.com> wrote:
> Hello,
>
> I am attempting to use ABF to induce the translocation of a 20-base
> polynucleotide through a protein pore and am having some trouble getting the
> system to work in the way that I was expecting. I have tested the tutorial
> files with success but have not been able to apply the technique to my
> system.
>
> I realise that it is difficult for others to comment on using ABF in an
> unfamiliar system, so listed below is what I believe to be the key
> information for consideration:
>
> The alpha carbons of the protein pore are constrained to maintain their
> positions.
> One set of alpha carbons belonging to a protein pore residue is used as
> "abf1" for the reference point to "abf2"s translocation. Given that abf1's
> position is constrained (see above point), I presume that is will allow
> consistent translocation of abf2.
> The atom I wish to translocate is the leading atom of the polynucleotide
> (thus pulling the whole molecule with it).
> The precise z-axis separation of abf1 and abf2 begins as -34.4900016784
> Angstroms.
> I wish to translocate the polynucleotide molecule from -34.5 to -42.5
> separation.
>
> The problem is that no matter what alterations I seem to make to the input
> parameters, the abf2 atom simply moves very quickly from xiMin to the middle
> of xiMin and xiMax, and remains within roughly +/- 0.3 Angstroms of the
> precise middle of those two values for the remainder of the simulation. If I
> have understood ABF correctly, I believe that the ABF scripts should sample
> the system while the abf2 atom resides within each dxi bin starting from
> xiMin and ending with xiMax, and once each dxi bin has been sampled "abf
> fullSamples" number of times, the adaptive biasing force will allow abf2 to
> overcome energy barriers in order to move to the next dxi bin, though the
> movement is also dependent on the abf2 atom's self diffusion properties.
>
> The fact that it so quickly moves to the middle of xiMin and xiMax shows
> that I have either set up the simulation incorrectly, that I have
> misunderstood what is supposed to occur during the simulation, or that my
> system is not compatible with the ABF technique. Any clues as to which of
> these three may be the case would be greatly appreciated.
>
> Listed below is an example of my ABF input parameters in the conf file,
> though I have also run test simulations with altered parameters, in
> particular "abf xiMin", "abf xiMax", "abf dxi", "abf fullSamples", and
> "numsteps" in order to try and resolve the issue:
>
> # ABF SECTION
>
> source               ../abf-1.8/abf.tcl
>
> # ORDER PARAMETER
>
> abf coordinate       zCoord-1atom
>
> # abf1 atoms are the CA atoms of protein resid 8, abf2 is the pulling atom
> abf abf1 {122863 95173 99788 104403 109018 113633 118248}
> abf abf2 262093
>
> # precise xiMin is -34.4900016784
> abf xiMin            -34.5
> abf xiMax            -42.5
>
> abf dxi              0.4
>
> # SAMPLING
>
> abf fullSamples      100
> abf dSmooth          0.1
>
> abf forceConst       10.0
>
> # OUTPUT
>
> abf writeXiFreq      200
> abf outFile          abf_window1.abf
>
> # MD SETUP
>
> numsteps                600000
>
>
>
> Many thanks,
>
>
> Hugh

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:51:10 CST