AMBER Archive (2002)

Subject: Re: Strange MD simulation results with GB solvation at high temp.

From: Soonmin Jang (sjang_at_protege.snu.ac.kr)
Date: Fri Sep 13 2002 - 00:21:32 CDT


I runned again with large cutoff (999.0 !) and the results are essentially
same. Still I see sudden temperature/potential energy fluctuation.
As I have mentioned in previous mail, the potential energy distribution
is not Boltzman anymore after this.

Please take a look my web page for more detailed graph if interested.

http://protege.snu.ac.kr/research/temp-met.pd
http://protege.snu.ac.kr/research/ene-met.pdf

David mentioned that there could be statistical artifact in this small
system. I agree with that but this is not so small system I guess.
Even though the system is small, there sould be consitant fluctuation
of the temperatre/enregy NOT sudden jump from small to large fluctuation.

I have not observed this kind of behavior in vacuume(the same system
with same temperature).
If you watch my input carefully, you will noticed that I am not using
periodic box condition (i.e. ntb=0).

Thank you.

Soonmin

On Wed, 11 Sep 2002, Carlos Simmerling wrote:

> As a side note, we've been looking atthe effect of using a cutoff
> (in proteins, AMBER7 with GB solvation). We find essentially no
> correlation between relative energies of a set of structures without
> cutoff and with cutoffs ranging anywhere from 8A to 20A
> (using the same set of structures in each case). We are working on a
> fix for this but the strongly suggest not using a cutoff with GB.
>
> Carlos
>
> From: "David A. Case" <case_at_scripps.edu>
> > I don't know if this is related to your problem, but using the default
> cutoff
> > (8 Ang.) is only recommended for PME calculations. Certainly a longer
> cutoff
> > would be needed for good simulations with gb.
>

--