Re: Calculation of Diffusion coefficient

From: Giacomo Fiorin (giacomo.fiorin_at_gmail.com)
Date: Mon Jan 29 2018 - 10:43:27 CST

On Mon, Jan 29, 2018 at 11:35 AM, Souvik Sinha <souvik.sinha893_at_gmail.com>
wrote:

> Okay, I get that calculating diffusion coefficient may not be of much use
> in this context.
>
> Sorry, I didn't get what you meant by this: "How about you simply take the
> last part of the ABF simulation, where the biasing force doesn't change
> much due to convergence, and you assume a net total force of zero on the
> variable? "
>

That's for estimating the diffusion coefficient.

>
> If the PMF is converged, then certainly the net total force would be zero
> on my specified variable. Sorry, I didn't check the convergence till now
> which I can definitely do by plotting the PMF at the different time point
> and see whether it is converged or not.
>

I'd recommend doing that: permeation of a whole peptide across a bilayer is
a difficult computation.

> But if I wanna check the uncertainty of the calculated PMF, I need to have
> the system force trajectory file for the calculation of say, mean
> autocorrelation of force at the different regime of the RC [ J. Phys.
> Chem. B, 2015, 119 (3), pp 1129–1151 ]. So, is there any way I can get
> that now without re-running all the windows?
>

For estimating the uncertainty of the PMF (unrelated to the diffusion
coefficient), try recomputing forces, Colvars and ABF on the existing DCD
files through "coorfile" and "run 0". If the result is not acceptable to
you, you may have to run additional trajectories, starting from the latest
restarts.

Giacomo

>
> Please excuse me being naive about all the intricacies.
>
> Thanks for your quick response.
>
> On Mon, Jan 29, 2018 at 9:32 PM, Giacomo Fiorin <giacomo.fiorin_at_gmail.com>
> wrote:
>
>> Hi Souvik, FYI you may also want to take a look at the more recent papers
>> by Jeff Comer regarding computing permeabilities of small molecules across
>> a bilayer.
>>
>> For something larger, like a peptide, you are definitely much more bound
>> by the PMF than by the differences in diffusion coefficient. This may be
>> less of a problem for you, because the diffusion coefficient changes by
>> less than an order of magnitude between lipid bilayers and bulk water.
>>
>> As for recomputing the system force along the trajectory, you do have its
>> average already computed along the collective variable, given by the PMF
>> (is it converged?). Computing the system force as a function of time would
>> only help you in collecting diffusion data during the transient regime
>> before convergence. Re-running on the single-precision, sparsely collected
>> trajectories may not be of much help.
>>
>> How about you simply take the last part of the ABF simulation, where the
>> biasing force doesn't change much due to convergence, and you assume a net
>> total force of zero on the variable?
>>
>> Giacomo
>>
>>
>>
>> On Mon, Jan 29, 2018 at 9:01 AM, Souvik Sinha <souvik.sinha893_at_gmail.com>
>> wrote:
>>
>>> Hi,
>>>
>>> Currently, I am estimating a peptide insertion (into membrane) free
>>> energy profile using ABF. I want to calculate the local diffusion
>>> coefficient along the reaction coordinate for which I need the estimation
>>> of 'time autocorrelation function of the fluctuations of the instantaneous
>>> forces' *[ J. Phys. Chem. B, 2011, 115 (49), pp 14704–14712 ]*.
>>> Unfortunately, I forgot to use the "*outputSystemForce*" option in the
>>> *'colvar' *block during the ABF simulations.
>>>
>>> From a previous thread (*http://www.ks.uiuc.edu/Research/namd/mailing_list/namd-l.2010-2011/2154.html
>>> <http://www.ks.uiuc.edu/Research/namd/mailing_list/namd-l.2010-2011/2154.html>*),
>>> I found that taking the assumption of similar force distribution and
>>> autocorrelation time across different window will certainly narrow the
>>> rerun option to a single window along a specific RC (for the purpose of
>>> error analysis). But I think, the autocorrelation function of instantaneous
>>> force will differ along RC because of the different environment (water to
>>> lipid bilayer) and that is what, I guess, will be reflected in the values
>>> of diffusion coefficient along RC. In that case, do I need to rerun all the
>>> windows with "*outputSystemForce*" option enabled (which is pretty
>>> troublesome right now)? Is there any post-processing option from which I
>>> can extract the system force trajectory for all the windows?
>>>
>>> Thanks.
>>> --
>>> Souvik Sinha
>>> Research Fellow
>>> Bioinformatics Centre (SGD LAB)
>>> Bose Institute
>>>
>>> Contact: 033 25693275
>>>
>>
>>
>>
>> --
>> Giacomo Fiorin
>> Associate Professor of Research, Temple University, Philadelphia, PA
>> Contractor, National Institutes of Health, Bethesda, MD
>> http://goo.gl/Q3TBQU
>> https://github.com/giacomofiorin
>>
>
>
>
> --
> Souvik Sinha
> Research Fellow
> Bioinformatics Centre (SGD LAB)
> Bose Institute
>
> Contact: 033 25693275
>

-- 
Giacomo Fiorin
Associate Professor of Research, Temple University, Philadelphia, PA
Contractor, National Institutes of Health, Bethesda, MD
http://goo.gl/Q3TBQU
https://github.com/giacomofiorin

This archive was generated by hypermail 2.1.6 : Tue Dec 10 2019 - 23:19:32 CST