Dear Madhur,
The end-state coordinate update in the cuda NEB code happens from step 1 to 2 in the loop of the molecular dynamics. Afterward, the end-state replicas remain fixed. I was aware of this behavior when I was writing the code, but I did not invest much time on it (this has been sitting in my TODO list for some time!). However, updating the coordinates once does not do any damage. The results you would get from performing NEB would still be correct and statistically equivalent to the previous implementation.
So, the reason your simulations are failing is not that. I believe if you pick a smaller spring constant (skmin, skmax = 1) at the initial stage where you superimpose the NEB replicas, your simulations would finish successfully.
Let me know if you were able to make it work.
-Delaram
________________________________
From: Madhur Aggarwal <madhur.aggarwal.research.iiit.ac.in>
Sent: Monday, November 18, 2019 2:44 PM
To: AMBER Mailing List
Cc: Ghoreishi, Delaram; Carlos Simmerling
Subject: Re: [AMBER] Conformation of NEB end states changing during 1HEAT
[External Email]
Hi Delaram, Carlos,
Thank you for your prompt reply.
As per Carlos's suggestion, I am focussing on NEB calcs in implicit solvent only for Alanine dipeptide (attached ala_implicit.1heat.sh) and my system for now ..
I compared the 2 mdin scripts line by line and I realized that there was one difference .. I had put SHAKE constraints on hydrogen atoms in my system, which were not there in case of Alanine dipeptide. I removed these constraints (attached modified implicit.1heat.sh) and ran simulation once again on my system.
For my system, by looking at the mdout file (attached mysys.neb.out.000), I see that the energies for end states change by a very small amount (in decimals) when their values are compared at NSTEPS=0 and NSTEPS=500 .. energy values for the end states remain constant after that ..
Whereas in the case of Alanine dipeptide, the energy values seem to remain constant throughout, upto 3 decimal places. (attached ala.neb.out.000).
When I compare the PDB files of the initial and final conformations of each end state, the coordinates differ. (Attached initial and final pdb files for one of my end state state1 here).
.Delaram, I have attached state1.prmtop, state1.inpcrd (after minimization) and state2.inpcrd (after minimization) files here.
Thanks a lot for your time,
Madhur Aggarwal
________________________________
From: Madhur Aggarwal <madhur.aggarwal.research.iiit.ac.in>
Sent: Tuesday, November 19, 2019 1:10 AM
To: AMBER Mailing List <amber.ambermd.org>; Ghoreishi, Delaram <delaram.phys.ufl.edu>; Carlos Simmerling <carlos.simmerling.gmail.com>
Subject: Re: [AMBER] Conformation of NEB end states changing during 1HEAT
Hi Delaram, Carlos,
Thank you for your prompt reply.
As per Carlos's suggestion, I am focussing on NEB calcs in implicit solvent only for Alanine dipeptide (attached ala_implicit.1heat.sh) and my system for now ..
I compared the 2 mdin scripts line by line and I realized that there was one difference .. I had put SHAKE constraints on hydrogen atoms in my system, which were not there in case of Alanine dipeptide. I removed these constraints (attached modified implicit.1heat.sh) and ran simulation once again on my system.
For my system, by looking at the mdout file (attached mysys.neb.out.000), I see that the energies for end states change by a very small amount (in decimals) when their values are compared at NSTEPS=0 and NSTEPS=500 .. energy values for the end states remain constant after that ..
Whereas in the case of Alanine dipeptide, the energy values seem to remain constant throughout, upto 3 decimal places. (attached ala.neb.out.000).
When I compare the PDB files of the initial and final conformations of each end state, the coordinates differ. (Attached initial and final pdb files for one of my end state state1 here).
.Delaram, I have attached state1.prmtop, state1.inpcrd (after minimization) and state2.inpcrd (after minimization) files here.
Thanks a lot for your time,
Madhur Aggarwal
________________________________
From: Carlos Simmerling <carlos.simmerling.gmail.com>
Sent: Monday, November 18, 2019 11:10 PM
To: AMBER Mailing List <amber.ambermd.org>
Subject: Re: [AMBER] Conformation of NEB end states changing during 1HEAT
I'm a little confused about what's happening. you say that alanine
dipeptide works fine in implicit solvent, but your system does not work
fine in implicit solvent? if so, then for now let's ignore explicit solvent
since you have 2 setups to compare, and one works while the other does not.
did you compare line by line the input file for your system vs the one you
used for the dipeptide that worked? you didn't include the working example
here so we can't test that. you also want to compare the run scripts, not
just the mdin files. something must not be the same.
it can also help to carefully look at the mdout files and make sure it is
doing what you think it should. again, you have a working example and a
non-working example, so you can compare them.
On Mon, Nov 18, 2019 at 12:10 PM Madhur Aggarwal <
madhur.aggarwal.research.iiit.ac.in> wrote:
> Hi all,
>
> I am using Amber18 to run NEB simulations on my protein + water system. I
> minimized both the end states and made sure the prmtop file generated is
> the same for both the end states. Then, I am following Tutorial A5 on NEB
> calcs on Alanine dipeptide.
> The issue is when I run step 4 to generate images and heat the system, my
> end states are also undergoing a change in their conformation, which should
> not happen (attached explicit.1heat.sh).
> When I try doing the calculations on Alanine dipeptide in implicit solvent
> (as done in the tutorial), it works fine. So, I thought maybe it has
> something to do with my explicit solvent. But when I do the calculations on
> my system using implicit solvent, the conformations of my end states are
> still changing (attached implicit.1heat.sh).
> I am not able to figure out why this is happening. I would really
> appreciate if someone could help me out with this. I have attached my input
> scripts for both implicit and explicit solvent cases here.
>
> Thanks,
> Madhur Aggarwal
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber<https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.ambermd.org_mailman_listinfo_amber&d=DwMFAg&c=sJ6xIWYx-zLMB3EPkvcnVg&r=xYIhOCyjOURFT_QMDQYnvtt9HjOAlBnUCY7kJn1oQfM&m=f15-7Jv99DlPse-PLPgxeLS9YwV7kn4Z4f3v27n6twM&s=XAabMXOFF9COTgh_sNcRHHYN1CxGhX7QGKGON3PPa04&e=>
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber<
https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.ambermd.org_mailman_listinfo_amber&d=DwMFAg&c=sJ6xIWYx-zLMB3EPkvcnVg&r=xYIhOCyjOURFT_QMDQYnvtt9HjOAlBnUCY7kJn1oQfM&m=f15-7Jv99DlPse-PLPgxeLS9YwV7kn4Z4f3v27n6twM&s=XAabMXOFF9COTgh_sNcRHHYN1CxGhX7QGKGON3PPa04&e=>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Nov 18 2019 - 23:00:03 PST
