Re: inconsistency in ABF results for helix unfolding

From: JC Gumbart (gumbart_at_ks.uiuc.edu)
Date: Wed Jan 28 2015 - 15:40:30 CST

Just to add to this, Chris Chipot, my student, and I needed two coordinates to characterize deca-alanine in water: http://pubs.acs.org/doi/abs/10.1021/ct5002076

On Jan 28, 2015, at 5:52 PM, Giacomo Fiorin <giacomo.fiorin_at_gmail.com> wrote:

> There is no universally defined reaction coordinate for the unfolding of an alpha-helix.
>
> FYI in one of the first applications of the colvars module (Henin et al, JCTC 2009) we used three variables to really nail deca alanine in vacuo.
>
> On Wed, Jan 28, 2015 at 3:50 PM, MEHRAN MB <mb.mehran1_at_gmail.com> wrote:
> Thank you so much for you quick reply,
>
> Niklaus, I totally agreed that shorter end to end distance does not necessary mean that the peptide is helica. Dihedral angles as a CV, only have four groups of atoms available regarding one single torsion angle, which means, I guess, 56 torsion angles need to be defined to cover all phi and psi dihedrals in 28 peptide chain?!?! Also I am not sure what should I considered as a reaction coordinates ? it is the same with h-bond colvar, there are many reaction coordinates correlated together!! I have no experience with this Colvar, so sorry if it is a naive question.
>
> ps: I ran 50ns for 5A windows (40 - 45) still not getting results
>
> Ajasja, I have not read that and I took a quick look, it should be a great starting point to see if I can resolve my issue. Hopefully there are some NAMD friendly solution in this paper.
>
>
> Thank you again, and I will keep you posted if I find any solutions.
>
>
> Best,
> Mehran
>
> On Wed, Jan 28, 2015 at 5:05 AM, Ajasja Ljubetič <ajasja.ljubetic_at_gmail.com> wrote:
> Exactly, choosing the colvar(s) is crucial. Meheran, have you perhaps read Comer et. al [1], in particular the section "Hidden Barriers and Other Challenges to Obtaining Accurate Results"?
>
> Best regards,
> Ajasja
>
> 1.
> Comer, J. et al. The Adaptive Biasing Force Method: Everything You Always Wanted To Know but Were Afraid To Ask. J. Phys. Chem. B (2014). doi:10.1021/jp506633n
>
>
> On 28 January 2015 at 10:49, gmail <niki.johner_at_gmail.com> wrote:
> Hi,
>
> I think that in free energy calculations, choosing a good collective variable is usually more important than the method used (which more or less all suffer from the same problems). Sampling is obviously the second big issue.
>
> It seems you used end to end distance as a collective variable, which does not strike me as the best choice. The peptide could be unfolded and still have a short end to end distance. On the other hand, a shorter end to end distance does not mean that the peptide is helical. And a long end to end distance constrains the peptide to be straight, hence largely limiting the conformational space available to the unfolded peptide.
>
> In that respect your test case might not be a great one, as alanine is known to show a strong bias towards helical conformations. Hence in that particular case the helical structure will be dominant for shorter distances whereas this might not be the case for your peptide. I would think about trying some other CV that would describe the end states of interest better (such as H-bonds and dihedral angles or the AlphaRMSD variable available in PLUMED).
>
> How long were your simulations? A 24 residue peptide has a much more conformational freedom than a 10 residue peptide, especially when unfolded. So much larger variations in the calculated free energy should be expected due to sampling issues.
>
> Best,
>
> Niklaus
>
> Dr. Niklaus Johner
> University of Basel
> Klingelbergstrasse 50 / 70
> CH - 4056 Basel
> Tél: 076 302 12 20
>
>
> On 28 Jan 2015, at 00:09, MEHRAN MB <mb.mehran1_at_gmail.com> wrote:
>
>> Dear NAMD experts,
>>
>> I am trying to compare free energy of forming helix structure for two coils with 28 residues in length. Both are identical except, in one all GLU are substituded by ALA. Indeed, I want to measure these residue contribution in forming helix, and I need to calculate Free energy with high precision.
>>
>> I tried ABF method and it works great for deca-alanine in vacuum. I run multiple simulations and all gave me pretty consistence results, both in curve shape and final value. free-energy vs reaction-coordinates result is provided. (Also when I solvate them in water I could not get good result)
>>
>> <deca-alanine_vacuum.png>
>>
>> However trying ABF on my 28 res polypeptide in vacuum, set boundary on end-to-end distance to vary from 40A to 80A and #samples within 40000 to 200,000; I have seen pretty inconsistency in free energy curve and final value. Since there are few charged residues I thought it might trapped in a non-equilibrium state.
>>
>> I divided the ABF process to smaller windows. I tried first small window, end-to-end distance varying from 40A to 45A and sampling from 100,000 to 200,000, and I am still getting pretty inconsistent result ( from 5 to 10 Kcal/mol ) for this window.
>>
>> So I believe I have enough sampling but it does not necessary mean that It has been sampled in equilibrium state. So How I can make sure that I am sampling in quasi-equilibrium process ?
>>
>> Next step would be measuring this helix unfolding Free energy in solution since I doubt I can say anything about the free energy in solution from vacuum simulation results. (I might be wrong, )
>>
>> any advice help to improve ABF results or suggestion about other methods will be appreciated,
>>
>>
>> Best,
>> Mehran
>> UOttawa
>>
>> here is simulation details:
>> 0.5 timesteps
>> rigidbond none
>>
>> 0.1 bin width
>> 1000 fullsampling
>
>
>
>
>
>
> --
> Giacomo Fiorin
> Assistant 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
> https://icms.cst.temple.edu/members.html
> http://giacomofiorin.github.io/
>

This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:21:36 CST