Topic Options
#6099 - 04/07/05 05:18 PM temperature control for implicit solvent simulation
ysun Offline
Forum Member

Registered: 05/06/04
Posts: 65
Loc: USA
Hi,
My protein system has about 1264 atoms. I run 200 ps using implicit solvent model with Nose-Hoover for temperature control at 298 K. The final portion of the output shown as bellow.

I saw the average temperature is around 298 K, which is good. However the RMS Fluctuations is about 7, is this fluctuation
considerred to be good or not for such a system?

---------- --------- --------- --------- --------- ---------
AVER> 500 206.00000 -4902.84149 1043.21749 -6100.42245 298.02494
AVER PROP> 20.73674 -5057.20495 0.00000 154.36346 972.83933
AVER INTERN> 337.86182 544.95146 0.00000 265.42629 169.19532
AVER EXTERN> -902.50197 -7334.51996 0.00000 0.00000 0.00000
AVER PRESS> 0.00000 -648.55955 0.00000 0.00000 0.00000
AVER ACE1> 819.16459 -6028.99800 4961.56097 -6267.08293
AVER ACE2> -1067.43703 -1305.52196
---------- --------- --------- --------- --------- ---------

* * * RMS FLUCTUATIONS FOR 500 STEPS
FLUC> 500 206.00000 1.47172 23.34342 28.05388 6.66872
FLUC PROP> 0.42584 36.67651 0.00000 36.65298 530.72636
FLUC INTERN> 15.68984 16.48215 0.00000 12.12114 9.70354
FLUC EXTERN> 15.24827 13.61741 0.00000 0.00000 0.00000
FLUC PRESS> 0.00000 353.81757 0.00000 0.00000 0.00000
FLUC ACE1> 4.05406 23.07827 34.45255 21.37864
FLUC ACE2> 21.46580 29.54444
---------- --------- --------- --------- --------- ---------

* * * AVERAGES FOR THE LAST 102500 STEPS
LAVE> 102500 206.00000 -4901.61646 1043.12999 -6051.37968 297.99994
LAVE PROP> 20.41132 -5008.24969 0.00000 106.63323 976.65860
LAVE INTERN> 330.80600 543.81468 0.00000 241.22367 163.99212
LAVE EXTERN> -894.18584 -7284.14000 0.00000 0.00000 0.00000
LAVE PRESS> 0.00000 -651.10573 0.00000 0.00000 0.00000
LAVE ACE1> 847.10970 -6244.11000 5173.72586 -6213.75586
LAVE ACE2> -1070.38414 -1040.03000

* * * RMS FLUCTUATIONS FOR 102500 STEPS
LFLC> 102500 206.00000 5.98738 24.50976 52.33771 7.00191
LFLC PROP> 0.41328 57.91018 0.00000 59.82191 557.52305
LFLC INTERN> 14.71067 17.62904 0.00000 12.72913 9.92702
LFLC EXTERN> 19.27877 40.95759 0.00000 0.00000 0.00000
LFLC PRESS> 0.00000 371.68203 0.00000 0.00000 0.00000
LFLC ACE1> 17.32407 111.78158 111.69518 94.08984
LFLC ACE2> 71.73882 146.42776
---------- --------- --------- --------- --------- ---------

Top
#6100 - 04/08/05 08:57 AM Re: temperature control for implicit solvent simulation [Re: ysun]
jb007 Offline
Forum Member

Registered: 09/30/03
Posts: 250
Loc: Florida State
Compare your RMSF wrt T to what other people report in the literature.

Top
#6101 - 04/08/05 09:25 AM Re: temperature control for implicit solvent simulation [Re: ysun]
feller Offline
Forum Member

Registered: 08/26/04
Posts: 5
It sounds as if it is working correctly. Extended system methods, such as Nose-Hoover, keep the mean temperature around the target. There is supposed to be a fluctuation. The magnitude of the fluctuation is a function of the heat capacity. Unfortunately, you won't be able to compare it quantitatively with anything from the literature unless they are using the exact same force field, etc., because the heat capacity (as measured from the fluctuation) will be a function of force field, shake vs no shake, and lots of other factors.

Top
#6102 - 04/08/05 05:02 PM Re: temperature control for implicit solvent simulation [Re: jb007]
ysun Offline
Forum Member

Registered: 05/06/04
Posts: 65
Loc: USA
Can you point out some literature I can look at ?

Thanks !

Yu

Top
#6103 - 04/08/05 06:03 PM Re: temperature control for implicit solvent simulation [Re: ysun]
rmv Online   content

Forum Member

Registered: 09/17/03
Posts: 8379
Loc: 39 03 48 N, 77 06 54 W
For implicit solvation with no friction, there's not much physical meaning for these thermal fluctuations. The magnitude is somewhat large because you have only a single independent particle, which probably means you'd have to run the simulation a lot longer for reasonable statistics in any event.

As an aside, the NOSE dynamics with the VVER integrator may not be a good choice for this. You'd be better off running Langevin dynamics, probably at low friction (SCALAR FBETA 2.0) if your main interest is conformational sampling. The Langevin bath gives you temperature control, and the low friction tends to give more dihedral transitions w/o affecting equlibrium distributions. Also, the VVER integrator is suspect, and has been replaced in newer CHARMM releases.
_________________________
Rick Venable
computational chemist


Top

Moderator:  John Legato, lennart