Re: ABF protocol to calculate PMF

From: Rukhsar Ali (arukhsar052_at_gmail.com)
Date: Wed Aug 12 2020 - 14:12:48 CDT

Many thanks for your suggestions.

On Thu, 13 Aug 2020, 12:06 am Giacomo Fiorin, <giacomo.fiorin_at_gmail.com>
wrote:

> Sorry, I truncated a sentence by accident. What I meant is that in this
> paper:
> https://doi.org/10.1016/j.bpj.2020.01.044
> I used multiple ABF windows, i.e. the same as suggested by Chris.
>
> My point is for this specific problem to invest more time in preparing the
> initial structures than deciding what sampling method to use.
>
> Giacomo
>
> On Wed, Aug 12, 2020 at 2:20 PM Giacomo Fiorin <giacomo.fiorin_at_gmail.com>
> wrote:
>
>> This is not a trivial problem, so no need to apologize. The problem of
>> hidden barriers is ubiquitous in PMF calculations and there is no general
>> solution to fix it. What you suggest in terms of comparing upper and lower
>> leaflet PMFs is reasonable, but ultimately you will have to check if it
>> works, and understand why it may not.
>>
>> One additional thing that I may suggest, as an alternative to using SMD
>> to initialize the windows, is trying to prepare the initial membrane
>> structure with the permeating molecule already embedded at the intended
>> depth for each window, and then equilibrate it. For a lipid bilayer with
>> exceptionally slow diffusivity I found this to give fairly accurate
>> permeabilities:
>> https://doi.org/10.1016/j.bpj.2020.01.044
>> compared to previous computations that used SMD to prepare the windows
>> before the equilibrium sampling. As for the sampling method, Ius
>>
>> One important thing that you need to keep in mind for a peptide is that
>> its internal structure and orientation may be just as important, or even
>> more so than the Z-depth. So whatever you do, I would spend some time in
>> comparing different approaches of preparing a few specific initial
>> structures, rather than go directly for a PMF computation across the whole
>> bilayer. A set of well-equilibrated unbiased MD simulations is always more
>> informative than an unconverged PMF computation using a bad reaction
>> coordinate.
>>
>> Giacomo
>>
>> On Wed, Aug 12, 2020 at 1:46 PM Rukhsar Ali <arukhsar052_at_gmail.com>
>> wrote:
>>
>>> Thank you for your email.
>>>
>>> I know that the Z-distance projection is the most well-used RC to
>>> calculate permeation events and I get that ABF is not an out-of-equilibrium
>>> pulling experiment. Still, there may have hidden barriers related to
>>> peptide orientation, secondary structures, formation of pores etc. This
>>> article https://pubs.acs.org/doi/abs/10.1021/jp506633n suggests that
>>> hysteresis calculation is a good indicator for hidden barriers.
>>>
>>> In these articles (by Tse et al), convergence was assessed by the
>>> hysteresis between upper and lower leaflet profiles.So, this should be
>>> similar to running ABF windows by taking alternate end states as initial
>>> structures.
>>>
>>> Sorry for being naive
>>>
>>> On Wed, 12 Aug 2020, 10:03 pm Chris Chipot, <chipot_at_illinois.edu> wrote:
>>>
>>>> Ali,
>>>>
>>>> it is not quite clear what you refer to backward and forward PMF—ABF is
>>>> run at thermodynamic equilibrium; it is not an out-of-equilibrium pulling
>>>> experiment. There are many published applications out there, whereby the
>>>> free energy was measured, using ABF, in the direction normal to a lipid
>>>> bilayer. These free-energy calculations are usually prefaced by steered MD
>>>> and long equilibration stages to prep the different windows along the
>>>> permeation pathway. You may want to look at Tse et al. JCTC 2018, 2019--000000000000dd5ca505acb2f9e3--

This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:09 CST