From: JC Gumbart (gumbart_at_physics.gatech.edu)
Date: Tue Aug 23 2016 - 22:12:48 CDT
Ah, good to see there is a Jacobian term contribution.
I’m not sure I follow your proposal though. I am using the differenceVector option, which means that the transformation is 0->1 by definition. If my reference structure is in the middle, one endpoint (+1) would be determined by the eigenvector - are you saying the second endpoint (-1) is driving in the opposite direction?
To be especially clear, I’m trying to find the difference in energy between two conformations, one of which is not a local minimum, at least in a free state.
Thanks!
JC
> On Aug 23, 2016, at 9:29 PM, Giacomo Fiorin <giacomo.fiorin_at_gmail.com> wrote:
>
> Hi JC, what you describe for RMSD is the result of the Jacobian term of the collective variable definition (handled explicitly by thermodynamic integration and ABF, implicitly by others). In the case of RMSD, the term can be quite significant and it yields a logarithmically diverging free energy profile at RMSD ~ 0.
>
> For eigenvector, there is also a Jacobian term that depends on the anisotropy of the various eigenvector components. See the Appendix of Jérôme's and my Mol Phys paper for the expressions of the two terms.
>
> I definitely don't think that the two endpoints are more accurately represented than any other.
>
> Aside from those considerations (which could be difficult to use in practice), have you considered setting the fitting/reference structure in between the two endpoints, so that the transformation goes from -1 to +1?
>
> Giacomo
>
>
>
> On Tue, Aug 23, 2016 at 9:02 PM, JC Gumbart <gumbart_at_physics.gatech.edu <mailto:gumbart_at_physics.gatech.edu>> wrote:
> Hi all,
>
> I used the eigenvector colvar for a simple transition between two close states of a protein (RMSD = ~2.6 A). One curious thing I observed is that the minimum falls at 0.5; in fact, I saw this in other runs as well, even using different starting structures. For something like RMSD, it seems obvious that there will be more conformations accessible the more you increase RMSD, which would lower the energy. I can’t convince myself if that is the case here, however. Certainly, it seems that each value of the colvar (in my case normalized from 0->1) does not dictate a unique structure, but is there a clear relationship with where one is on the path and the number of accessible states? In which case, only the end points can truly be trusted?
>
> Thanks!
> JC
>
>
>
>
> --
> Giacomo Fiorin
> Associate Professor of Research
> Institute for Computational Molecular Science (ICMS)
> College of Science and Technology, Temple University
> 1925 North 12th Street (035-07), Room 704D
> Philadelphia, PA 19122-1801
> Phone: +1-215-204-4213 <tel:%2B1-215-204-4213>
>
> Scholar: http://goo.gl/Q3TBQU <http://goo.gl/Q3TBQU>
> Personal: http://giacomofiorin.github.io/ <http://giacomofiorin.github.io/>
> Lab page: https://icms.cst.temple.edu/members.html <https://icms.cst.temple.edu/members.html>
>
> "As computer programmers we have a responsibility to make sure that we run the computers instead of the computers running us." - Steve Oualline
>
>
This archive was generated by hypermail 2.1.6 : Sun Dec 31 2017 - 23:20:40 CST