Thank you Prof Case for the helpful insights.
I know its been 3 months or more that I am trying to get this working.
I have tried the following changes till now to make the TI-work on GPUs.(pmemd.cuda)
I would have to eventually do pmemd.MPI but our lab has limited cpu processors and me switching to TI-on cpus would mean others would have no running jobs, also the project would take 1 yr or more to get data out. That’s the reason I m trying my best to get GPU-TI running. Sorry for asking tons of questions on this.
1. Minimized longer (50000 steps of SD) , and made my dx0=0.0001 , don’t know if that helps, because I am having no issues with minimizing my complex(Kinase+small ligand)
2. Changed taup=1 to taup=2 to help in pressure relaxation.
3. Changed gamma_ln=5 instead of gamma_ln=3…
4. Changed to MC barostat=2
5. Heating was eventually done from 0-50K, then 50-100K, 100-200K, 200-300K, instead of instaneous temp wrapping.
6. Equilibrate NVT (2ns) followed by NPT (5ns)
7. Changed the thermostat tried ntt=3, ntt=1…
8. Made ntpr=1 to see if that helps.
9. Now I could decrease my restrain_wt to 10 as you have suggested, but the problem is then the ethanol molecule would not remain tethered to the spot for my binding free energy calculation. I cannot let the molecule move a lot away from the hotspot for which I am estimating the absolute FE.
10. I think I have tried some more variants to make my equilibration work on GPUs, but no luck as yet.
11. I will try equilibrating the system longer before switching on my TI.
Thanks for being patient and for helpful suggestions.
Regards
Debarati
Sent from Mail<
https://go.microsoft.com/fwlink/?LinkId=550986> for Windows 10
From: David Case<mailto:david.case.rutgers.edu>
Sent: Tuesday, December 17, 2019 8:39 AM
To: AMBER Mailing List<mailto:amber.ambermd.org>
Subject: Re: [AMBER] NVT equilibriation followed by NPT in TI
On Mon, Dec 16, 2019, Debarati DasGupta wrote:
>Page 327 Section 17.6.8 says it’s a good practice to do “NVT
>equilibration followed by NPT”…” Please note: in general you will need to
>equilibrate the temperature to something like the final temperature using
>constant volume (ntp=0) before switching on constant pressure simulations
>to adjust the system to the correct density”
>In my equilibration input files for my TI project, I have not followed
>this, is this possibly the reason why my system experiences high forces
>and GPU cannot handle the equilibration steps:
It's certainly a good idea to try. I'd also suggest doing a complete
"equilibration" at both end points (before turning on TI). Run for
quite a while on the GPU and see if any problems arise.
I'm sorry you are having so much trouble with this. I'm also guessing
taht most people on the list (like me) have lost track of all of the
things you have tried. The key to debugging is to try to find the
minimal change that goes from working to non-working. Two additional
general suggestions:
don't set iwrap=1 (possible but unlikely culprit)
use a restraint_wt of 10 rather than 100 (reduces large forces)
Since it's possible that there is problem with the ligand force field
parameters, try using something GLY in it's place, just to (again) try
to narrow down the possible causes of the problem.
...good luck...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Dec 17 2019 - 07:00:03 PST
