Re: [AMBER] TI not compatible with GB on GPU

From: Debarati DasGupta <debarati_dasgupta.hotmail.com>
Date: Wed, 21 Oct 2020 14:46:07 +0000

This is the only tutorial I found for TI runs. Please correct me if I am wrong.
http://ambermd.org/tutorials/advanced/tutorial9/#setup_pmemd

Sent from Mail<https://go.microsoft.com/fwlink/?LinkId=550986> for Windows 10

From: Vaibhav Dixit<mailto:vaibhavadixit.gmail.com>
Sent: 21 October 2020 09:00
To: AMBER Mailing List<mailto:amber.ambermd.org>
Subject: Re: [AMBER] TI not compatible with GB on GPU

Dear Debarati,
I think pmemd.MPI doesn't look like a correct usage, shouldn't it be
pmemd.cuda.MPI? Moreover, it would apply only when the jobs are distributed
between two GPU cards. That's what I interpreted from here
<https://ambermd.org/GPUHowTo.php>.
I have only one GPU card on this machine thus I was trying to run a single
GPU job.
It would be great if you can share your protocol (sample scripts and
commands) in brief or share a tutorial that you are following.

I'm estimating free energy of electron transfer (ET) and not protein-ligand
binding free E, these ET free-Es are expected to between 95-110 kcal/mol.
I'm assuming that the pKa shift tutorial is analogous to my case, they got
~55-60 kcal/mol free-Es and thus I'm trying similar simulations for my
system which would be speeded up significantly on GPU. At present on HPC
with 16 CPU-cores I'm getting only 0.17 ns/day which is very slow.
Please let me know if you have an additional suggestion for me.
I hope developers on the list will soon respond and give valuable
suggestion on these TI simulations and their advantages/limitations and
cautions to be kept in mind.
Thank you and best regards.
Vaibhav




On Wed, Oct 21, 2020 at 5:52 PM Debarati DasGupta <
debarati_dasgupta.hotmail.com> wrote:

> Hi Vaibhav,
> Welcome to the TI club!
> I feel on GPUs, you should be using pmemd.MPI not pmemd.cuda as I have
> used pmemd.MPI while running my TI jobs.
> Also I am not sure what is the exact system your studying as 100-150
> kcal/mol looks like a very very high energy binding process.
> I have some weak binders I work on and my free energies are in the range
> of -1 to -3 kcal/mol for binding to my kinase.
> Also the Amber20 TI-GPU is pretty fast. On a single GPU, I get about
> ~70ns/day in equilibration stages.
> I am not sure if we could use implicit solvents for TI, I have been using
> TIP3P water.
> Thanks
> I hope I could help somewhat.
> Debarati
>
> Sent from Mail<https://go.microsoft.com/fwlink/?LinkId=550986> for
> Windows 10
>
> From: Vaibhav Dixit<mailto:vaibhavadixit.gmail.com>
> Sent: 20 October 2020 09:02
> To: AMBER Mailing List<mailto:amber.ambermd.org>
> Subject: [AMBER] TI not compatible with GB on GPU
>
> Dear Amber users/experts,
> I'm running at TI calculation on CPU but it is expected to take 353 hr to
> complete.
> When I tried to run the same on GPU with pmemd.cuda I'm getting hte subject
> error mentioned below.
> Can someone on the list please guide me if there is a way around this or is
> it not possible with Amber18 on GPU?
> Additionally, I was expecting free energy values in the range 100-115
> kcal/mol, but I'm getting 180 kcal/mol values. I did run a 30 ns
> simulation on the model compound but values did not change much. I used
> procedure similar to that given in the tutorial here
> <http://ambermd.org/tutorials/advanced/tutorial6/section4.php>. My system
> is Heme with and without extra electron, charges were derived from MCPB.py
> tool.
>
> We are processing the purchase of Amber20, but it might take 15-20 days or
> more before the payment goes through. Is this something available in
> Amber20?
> thank you and best regards.
>
> Error while running GB based TI on GPU
>
> Note: ig = -1. Setting random seed to 390103 based on wallclock time in
> microseconds.
> | irandom = 1, using AMBER's internal random number generator (default).
> | ERROR: TI is incompatible with GB on GPU
>
> Input errors occurred. Terminating execution.
>
> --
>
> Regards,
>
> Dr. Vaibhav A. Dixit,
>
> Visiting Scientist at the Manchester Institute of Biotechnology (MIB), The
> University of Manchester, 131 Princess Street, Manchester M1 7DN, UK.
> AND
> Assistant Professor,
> Department of Pharmacy,
> ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
> Birla Institute of Technology and Sciences Pilani (BITS-Pilani),
> VidyaVihar Campus, street number 41, Pilani, Rajasthan 333031.
> India.
> Phone No. +91 1596 255652, Mob. No. +91-7709129400,
> Email: vaibhav.dixit.pilani.bits-pilani.ac.in, vaibhavadixit.gmail.com
> http://www.bits-pilani.ac.in/pilani/vaibhavdixit/profile
> https://www.linkedin.com/in/vaibhav-dixit-b1a07a39/
>
> ORCID ID: https://orcid.org/0000-0003-4015-2941
>
> http://scholar.google.co.in/citations?user=X876BKcAAAAJ&hl=en&oi=sra
>
> P Please consider the environment before printing this e-mail
> _______________________________________________
> 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
>


--

Regards,

Dr. Vaibhav A. Dixit,

Visiting Scientist at the Manchester Institute of Biotechnology (MIB), The
University of Manchester, 131 Princess Street, Manchester M1 7DN, UK.
AND
Assistant Professor,
Department of Pharmacy,
▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
Birla Institute of Technology and Sciences Pilani (BITS-Pilani),
VidyaVihar Campus, street number 41, Pilani, Rajasthan 333031.
India.
Phone No. +91 1596 255652, Mob. No. +91-7709129400,
Email: vaibhav.dixit.pilani.bits-pilani.ac.in, vaibhavadixit.gmail.com
http://www.bits-pilani.ac.in/pilani/vaibhavdixit/profile
https://www.linkedin.com/in/vaibhav-dixit-b1a07a39/

ORCID ID: https://orcid.org/0000-0003-4015-2941

http://scholar.google.co.in/citations?user=X876BKcAAAAJ&hl=en&oi=sra

P Please consider the environment before printing this e-mail
_______________________________________________
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 Wed Oct 21 2020 - 08:00:02 PDT
Custom Search