Re: ABF - stretching deca-alanine

From: Jerome Henin (jhenin_at_cmm.chem.upenn.edu)
Date: Mon Oct 22 2007 - 09:22:40 CDT

Well, it's not easy to decide when ABF will be converged before you're
actually there.

You can estimate an order of magnitude for the time it will take to
explore the range of your coordinate of interest. That will be the
fullSamples parameter times the number of bins. In real life, though,
it also depends on the diffusion coefficient along the coordinate,
since ABF sampling is based on diffusion. And the diffusion
coefficient is related to relaxation times of other degrees of freedom
that are coupled to the coordinate of interest.

Concerning the time to reach convergence, that's a different story,
because evolution of the system has to be reversible. In the end, the
reaction coordinate should be able to diffuse *back and forth* at any
point on the free energy surface, and the sampling should be roughly
uniform.

To get a first idea, what you can do is look at the literature and try
to find comparable systems on which ABF was done. For large biological
systems, long sampling may be necessary (>100ns). Deca-alanine is a
relatively simple and cheap case, that's why we chose it as an
example.

Jerome

On 10/22/07, Per Jr. Greisen <pgreisen_at_gmail.com> wrote:
> Thanks, how to estimate the time for ABF to converge?
>
>
> On 10/22/07, Jerome Henin <jhenin_at_cmm.chem.upenn.edu > wrote:
> > Hi,
> > Apparently, you have run approximately ten thousand timesteps. That's
> > too short to sample the stretching of the helix the ABF way, that is,
> > reversibly. You need on the order of a nanosecond to start having a
> > broad sampling of the alpha-to-extended transition.
> >
> > Best,
> > Jerome
> >
> >
> > On 10/22/07, Per Jr. Greisen <pgreisen_at_gmail.com> wrote:
> > > Hi,
> > >
> > > In the NAMD documentation for ABF, there is an example for stretching
> > > deca-alanine. If I apply the input to the configuration file - the
> alanine
> > > is not stretched as I would expected from ordinary application of force.
> The
> > > ABF gives me the following result:
> > > 12.100 0.0000 0.0000 0
> > > 12.300 0.0000 0.0000 0
> > > 12.500 0.0000 0.0000 0
> > > 12.700 0.0000 0.0000 0
> > > 12.900 -0.5639 5.6388 26
> > > 13.100 -1.1912 0.6341 121
> > > 13.300 -1.5174 2.6284 139
> > > 13.500 -1.9330 1.5278 258
> > > 13.700 -2.3387 2.5289 555
> > > 13.900 -2.7024 1.1075 1115
> > > 14.100 -2.8608 0.4774 987
> > > 14.300 - 2.7992 -1.0937 978
> > > 14.500 -2.6325 -0.5737 957
> > > 14.700 -2.4741 -1.0104 1174
> > > 14.900 -2.2683 - 1.0472 1171
> > > 15.100 -2.0278 -1.3581 930
> > > 15.300 -1.7928 -0.9914 722
> > > 15.500 -1.5389 -1.5477 529
> > > 15.700 -1.2097 -1.7441 257
> > > 15.900 -0.4726 -5.6270 81
> > > 16.100 0.0901 0.0000 0
> > > 16.300 0.0901 0.0000 0
> > > 16.500 0.0901 0.0000 0
> > > 16.700 0.0901 0.0000 0
> > > 16.900 0.0901 0.0000 0
> > > 17.100 0.0901 0.0000 0
> > > 17.300 0.0901 0.0000 0
> > > 17.500 0.0901 0.0000 0
> > > 17.700 0.0901 0.0000 0
> > > 17.900 0.0901 0.0000 0
> > > 18.100 0.0901 0.0000 0
> > > 18.300 0.0901 0.0000 0
> > > etc etc
> > > Can one stretch fx. a peptide using the ABF module? Thanks in advance.
> > > Best regards
> > > Per
> > >
> > >
> >
>
>
>
> --
> Best regards
> Per Jr. Greisen
>
> "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 - 15:45:24 CST