Re: [AMBER] force-momentum based SGLD algorithm

From: Wu, Xiongwu (NIH/NHLBI) [E] <"Wu,>
Date: Thu, 4 Jun 2015 15:14:51 +0000

Hi Barbara,

Sorry I missed your previous posting. For SGLDfp(isgld=2), the enhancement in sampling is limited to maintaining the ensemble distribution and may not be able to reach the input TEMPSG. It is safe to specify SGFT and let SGLDfp to reach a TEMPSGI (see output) while maintaining the ensemble distribution. The output TEMPSGI shows the searching ability in a temperature scale.
Please note that SGLDfp need additional input: treflf, for accurate ensemble distribution.
Alternatively, AMBER14 has option isgld=3 that provides SGLDg method to maintain canonical ensemble.
Maintaining canonical ensemble (isgld=2 or isgld=3) always reduce the enhancement in conformational sampling as compared with SGLD (isgld=1).
Please let me know if you have any further question.
Xiongwu Wu

> From: barbara rossi <>
> To: "" <>
> Date: Wed, 3 Jun 2015 11:22:16 +0200
> Subject: [AMBER] force-momentum based SGLD algorithm
> Hello all, I try to re-post my doubt one time more:
> I am running self-guided langevin dynamics with Amber12 and I decided
> to try the force-momentum based algorithm
> (isgld=2)
> The output contains the following warning:
> Guiding range from 1 TO 762
> Local averaging time: 0.2000 ps
> Guiding temperature: 380.00 K
> *** WARNING: SGFT, instead of TEMPSG, should be set for SGMDfp/SGLDfp!
> TEMPSG defines a seaching ability comparable to a simulation at
> temperature TEMPSG
> My question is: is the SGLDfp algorithm not working properly by
> specifying tempsg rather than sgft?
> Is it true also with isgld=2 that "TEMPSG defines a seaching ability
> comparable to a simulation at temperature TEMPSG" as it is the case
> for isgld=1?
> any experience?,
> thanks in advance.
> Barbara Rossi
> University of Liege
> Belgium
> _______________________________________________
> AMBER mailing list

AMBER mailing list
Received on Thu Jun 04 2015 - 09:00:06 PDT
Custom Search