Re: [AMBER] Help in Standardizing NEB

From: Aravind R via AMBER <amber.ambermd.org>
Date: Wed, 26 Oct 2022 16:11:22 +0530

Hi Christina,
Thanks for the detailed reply!

> NEB is a path sampling method that's used to examine the transition
> between two well-defined endpoints. It does not impose a barrier - the
> system's transition over this barrier is the limiting factor in sampling
> the open-to-close transition, and this is what/why you are trying to
> calculate the path (and potential energy) independent of timescale.
>

I am trying to understand the interactions governing the open-to-close
transition. Also, I have two test cases, where the Kon and Koff of these
transitions are severely perturbed/changed. Once I understand one set of
systems, I would like to compare it to the intermediate structure across
the transition and transition energy. While I understand that Barrier
height is not "imposed", I see the energy of the system increase while
transitions and the corresponding structures from the trajectory show the
complete unfolding of the helix (Structures attached). This "unfolding" is
constant with the various number of images/ spring constant ranges
attempted. This gave me the impression that Higher energy was forced.

I do have a few additional suggestions, which may help:
> -Try explicit solvent.
>

I was planning to do that once the protocol is standardised. But yes, I
will continue with explicit solvent NEB.

> -Extend the initial 1HEAT step to be more gradual (Try 1ns? 40ps is short,
> and might be ok for alanine dipeptide but not for everything).
>
-Add an equilibration step after the first heating step (500 ps, 1 ns?)
>

I will try these. Also, the unfolding of the helix is mostly introduced
while annealing (at 500K)

> -Add images at the intermediate to try and avoid discontinuities in the
> transition, this could help avoid those initial 'NaN' values.
>

I have tried this with various intermediates from TMD.

Look at Test Case 3 in this paper and change up the original inputs to
> match what they do (a similarly large system's protocol could be more
> relevant to your own problem):
> https://pubs.acs.org/doi/10.1021/acs.jctc.9b00329
>
Thanks for this suggestion. I will modify my protocol similar to that and
try to get a good interpolating initial path.

Regards,
Aravind R

>
> On Fri, Oct 21, 2022 at 6:43 PM Christina Bergonzo <cbergonzo.gmail.com>
> wrote:
>
>> Hi Aravind,
>>
>> NEB is a path sampling method that's used to examine the transition
>> between two well-defined endpoints. It does not impose a barrier - the
>> system's transition over this barrier is the limiting factor in sampling
>> the open-to-close transition, and this is what/why you are trying to
>> calculate the path (and potential energy) independent of timescale.
>>
>> It seems like you're doing some stuff to try and get this working for
>> your system, like increasing/changing the number of starting images,
>> generating intermediates using TMD, etc.
>> There is no "standard" way to run these simulations, unfortunately. Each
>> system is different, each underlying potential energy surface is different.
>>
>> I do have a few additional suggestions, which may help:
>> -Try explicit solvent.
>> -Extend the initial 1HEAT step to be more gradual (Try 1ns? 40ps is
>> short, and might be ok for alanine dipeptide but not for everything).
>> -Add an equilibration step after the first heating step (500 ps, 1 ns?)
>> -Add images at the intermediate to try and avoid discontinuities in the
>> transition, this could help avoid those initial 'NaN' values.
>>
>> Look at Test Case 3 in this paper and change up the original inputs to
>> match what they do (a similarly large system's protocol could be more
>> relevant to your own problem):
>> https://pubs.acs.org/doi/10.1021/acs.jctc.9b00329
>>
>> The real struggle is getting a nice interpolating initial path. I'd
>> recommend spending time on that first, then troubleshooting any other
>> problems that arise.
>>
>> Good luck!
>> -Christina
>>
>> On Fri, Oct 21, 2022 at 2:18 AM Aravind R via AMBER <amber.ambermd.org>
>> wrote:
>>
>>> Hi Amber Developers/ Users,
>>> I am performing NEB, moving the system from its "Open" structure to the
>>> "Close" structure. This comprises of large conformational transition.
>>> I had minimised the initial and final structures until convergence before
>>> starting NEB.
>>>
>>> I have played around with spring constants ( skmin/skmax ranging from 0.5
>>> to 50 at various stages), temperatures (slow to fast heating & slow
>>> cooling) and the number of images to represent this transition (I
>>> performed
>>> TMD to sample images from open to close transition. Used conformations
>>> from
>>> TMD to represent the number of images I supply NEB to start with - I
>>> used 8
>>> - 64 images to represent transition with just 2 to 32 conformations from
>>> TMD).
>>>
>>> I performed implicit solvent NEB in 6 stages:
>>> 1HEAT - spring constant
>>> ranging from 0.5 - 10 for 40ps of MD with a 0.5fs time step raising temp
>>> from 0 to 300K
>>> 2ANNEAL - spring constant
>>> ranging from 1 - 50 for 1ns of MD with a 1fs time step raising temp from
>>> 300 to 500K
>>> 3EQUIL - spring constant
>>> ranging from 1 - 50 for 2ns of MD with a 2fs time step equilibrating at
>>> 500K
>>> 4COOL - spring constant
>>> ranging from 1 - 50 for 2ns of MD with a 2fs time step decreasing temp
>>> from
>>> 500 to 300K
>>> 5EQUIL - spring constant
>>> ranging from 1 - 50 for 2ns of MD with a 2fs time step equilibrating at
>>> 300K
>>> 6COOL - spring constant
>>> ranging from 1 - 50 for 2ns of MD with a 2fs time step decreasing temp
>>> from
>>> 300 to 0K
>>> 6EQUIL - spring constant
>>> ranging from 1 - 50 for 2ns of MD with a 2fs time step equilibrating at
>>> 0K
>>> **spring constant ranging from x - y: is across different attempts not
>>> in a
>>> single simulation.
>>> I am having the following issues:
>>> 1) Irrespective of the values I play around with, NEB constantly
>>> imposes a barrier height. This leads to the unfolding of the helix in the
>>> region I try to perform the tgtrmsmask in.
>>>
>>> 2) At the images Where the conformational transition is sampled, the
>>> energies go very high and the potential energy gets to NaN (This mostly
>>> happens due to the helix unfolding raising dihedral and vdw energies to
>>> shoot up).
>>>
>>> 3) Even though I used 32 evenly spaced conformers (sampling the
>>> transition) from TMD to sample 64 images in NEB, the transition is not
>>> smooth, with a sudden drop in energy for the last but one conformer. This
>>> drop in energy moves towards the left as I decrease the number of
>>> conformers I use to represent transition. For eg: If I just use open and
>>> closed conformers (as in the tutorial), the dip in the energy is in the
>>> centre image. As I increase the Conformers from TMD the dip move towards
>>> the right from the centre. To negate this I tried an unequal number of
>>> images used to represent transition: the first 4 images were sampled from
>>> "Open" and the rest from "Close", and so on and so forth.
>>>
>>> Any help is appreciated !!!
>>>
>>> Regards,
>>> Aravind R
>>> _______________________________________________
>>> AMBER mailing list
>>> AMBER.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>
>>
>>
>> --
>> -----------------------------------------------------------------
>> Christina Bergonzo
>> Research Chemist
>> Biomolecular Measurement Division, MML, NIST
>> -----------------------------------------------------------------
>>
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Oct 26 2022 - 04:00:05 PDT
Custom Search