Re: Some questions on the Deca-Alanine ABF tutorial

From: Jérôme Hénin (jerome.henin_at_ibpc.fr)
Date: Tue May 17 2016 - 10:33:03 CDT

Hi,

On 16 May 2016 at 11:19, sunyeping <sunyeping_at_aliyun.com> wrote:

> Dear all,
>
> I am studying the Deca-Alanine ABF tourial and ABF method. I have finished
> the 5 ns run of the tutorial but I have several questions on the analysis
> of the results. In the tutorial the figure 1 in page 10 gives the results
> of the potential (delta A(ξ))as a function of the distance (ξ).The range
> of the delta A(ξ)is from around -4 to 23 kcal/mol. The tourial says that
> convergence usually occurs within 5 ns and one of the criteria of
> convergence is that the pmf does not vary any more. It seems that we should
> exam the change of pmf against time to judge wether the simulation has
> converged. Is there a output file that gives the pmf vs time result?
>

That file is only saved if you set the historyFreq parameter.

> In my output pmf files, the delta A(ξ)ranges from about 0.2 to 39
> kcal/mol, which is different from the tutorial. Is this OK? Another
> question is, in the figure 1 in the tourial, the inset shows the number of
> samples at each distance. How is the number of samples counted? Is it
> counted from the traj files? However, there are only 10000 frames in the
> traj file, while the number of samples in the inset figure are more than 1
> ×105 at each distance. Where does these numbers come from?
>

The trajectory file has one value per colvarsTrajFrequency steps, whereas
there is one ABF sample per MD timestep.

> The tutorial claims that the unfolding of deca-alaline is reversible
> and the different conformational states of the peptide chain are visited frequently.
> However, in my simulation, the unfolded peptide revert to the helix only in
> the last frame? Is this correct?
>

This is correct on average and for long trajectories. You may see it by
running more trajectories with the same parameters.

> The NAMD user's guide says that reconstructing multidemensional free
> energy surface requires abf_integrate command. But I cannot find it in the
> NAMD binary packages. where is this command implemented?
>

It's provided as source code only. This is explained here:
http://www.ks.uiuc.edu/Research/namd/mailing_list/namd-l.2009-2010/3491.html

Best,
Jerome

This archive was generated by hypermail 2.1.6 : Sun Dec 31 2017 - 23:20:26 CST