Hi Andrea,
Although I intentionally wrote this part so that simply changing MaxTIAtoms would work, I am still a little bit surprised that it really works😊. Please keep us posted about your results.
Thanks!
Taisung
-----Original Message-----
From: Andrea Cavalli <andrea.cavalli.irb.usi.ch>
Sent: Wednesday, March 27, 2019 8:38 AM
To: AMBER Mailing List <amber.ambermd.org>
Subject: Re: [AMBER] TI atom limit
Dear Taisung.
Thanks for your answer.
I modified the MaxTIAtoms variable to 2000 and recompiled. Seems to work.
Scientifically… I hope it makes sense… Theoretically yes, but might not converge. Could be a good test.
I let you know.
Thanks
Andrea
----------------------------------------------------------------
Andrea Cavalli
Group Leader. Computational Structural Biology Institute for Research in Biomedicine Università della Svizzera Italiana Via Vincenzo Vela 6
CH-6500 Bellinzona
Phone : +41 91 820 0391
Fax: : +41 91 820 0305
> On 27 Mar 2019, at 12:53, <accuratefreeenergy.gmail.com> <accuratefreeenergy.gmail.com> wrote:
>
> Hi Andrea,
>
> The TI limit of 200 atoms was set based on the assumption that TI regions should only contain few atoms. The design of the whole TI-GPU part is based on this assumption.
>
> Scientifically, TI/FEP are only practically useful when the perturbation is small. I would not suggest to do something like 200-atom mutated to nothing. Even you could do this kind of simulations, I don't think you will get any meaningful results--likely the errors will be huge.
>
> If this is a community consensus, technically we could increase the limit but it would also impact the performance because some pre-defined arrays on GPU would require more memory,...etc.
>
> Taisung
>
> -----Original Message-----
> From: Andrea Cavalli <andrea.cavalli.irb.usi.ch>
> Sent: Wednesday, March 27, 2019 4:15 AM
> To: amber.ambermd.org
> Subject: [AMBER] TI atom limit
>
> Hello,
> I want to use TI to decouple a protein in a dimer and compute the absolute free energy of binding. Unfortunately there is a limit of 200 atoms.
> Is there a way to increase it? (By changing the code) Andrea
>
>
> ----------------------------------------------------------------
> Andrea Cavalli
> Group Leader. Computational Structural Biology Institute for Research
> in Biomedicine Università della Svizzera Italiana Via Vincenzo Vela 6
> CH-6500 Bellinzona
>
> Phone : +41 91 820 0391
> Fax: : +41 91 820 0305
>
>
>
>
> _______________________________________________
> 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
_______________________________________________
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 Mar 27 2019 - 06:30:04 PDT
