Re: inconsistence in NAMD and AMBER energy

From: Floris Buelens (floris_buelens_at_yahoo.com)
Date: Tue Jul 21 2009 - 02:11:51 CDT

Hi,

I'm not sure about exactly how AMBER handles this issue, but I suspect the difference might arise from the neglect of van der Waals interactions beyond the cutoff. I think AMBER applies an analytical correction by default for these interactions (search the manual for 'vdwmeth'), while NAMD as of 2.7b1 doesn't yet. I would suggest re-running your analysis using an increasing series of cutoff values, I think you should see the NAMD value converging towards that of AMBER. Alternatively, the option for analytical corrections for long-range van der Waals has now been added to the latest development code, which you can access through CVS as described on the NAMD website, if you use that code with the option "LJCorrection on" I hope you shouldn't see such a big discrepancy.
Best wishes,

Floris

________________________________
From: Yi Shang <mirandaisbest_at_gmail.com>
To: namd-l_at_ks.uiuc.edu
Sent: Tuesday, 21 July, 2009 1:31:47
Subject: namd-l: inconsistence in NAMD and AMBER energy

Hi all,
I did some comparison of energy calculated by NAMD 2.7b1 and AMBER 10. I found something intriguing.
During simulation, VDW energy from AMBER simulation is constantly much lower from NAMD. So I tried to compare point energy first. I turned off SHAKE, temperature, and pressure control and only looked at basic energy calculation of one single structure: bonds, angles, dihedrals, electrostatic, and vdw energies.
For non-periodic boundary, I had no problem, energies are exactly the same.
For periodic boundary condition here comes the problem:
-electro energy off by ~14kcal/mol, which is ok, parameters might not be identical in two methods.
-vdw energy off by > 600 kcal/mol. That's terrible.
 
I tested these inputs on two different periodic systems, and I get the same result (same energy gap). So I don't think it's system specific problem.
Here is my inputs and outputs from one system's comparison (I combined AMBER VDWAALS and 1-4 NB energies to compare to NAMD vdw energy, likewise for electrostatic interaction):
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
NAMD config.in:
amber on
parmfile box.top
ambercoor box.crd
binaryoutput no
outputname md6
outputpressure 1
outputenergies 1
exclude scaled1-4
1-4scaling 0.83333
timestep 2
temperature 0
cutoff 8
switching off
PME yes
PMEGridSpacing 1.0
cellbasisvector1 79.6614 0 0
cellbasisvector2 0 60.2065 0
cellbasisvector3 0 0 63.7403
cellorigin 0 0 0
run 0
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
amber md.in:
&cntrl
imin = 0, nstlim = 1, dt = 0.002,
irest = 0, ntx = 1,
tempi = 0.0, temp0 = 0.0,
ntwx = 0, ntwe = 0, ntwr = 0, ntpr = 1,
cut = 8.0,
/

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
namd energies:
ETITLE: TS BOND ANGLE DIHED IMPRP ELECT VDW BOUNDARY MISC
       KINETIC TOTAL TEMP POTENTIAL TOTAL3 TEMPAVG PRESSURE GPRESSURE VOL
UME PRESSAVG GPRESSAVG
ENERGY: 0 341.5078 716.0018 2372.7673 0.0000 -67739.0356 7208.9795 0.0000 0.0000
        0.0000 -57099.7791 0.0000 -57099.7791 -56529.2524 0.0000 11740.3005 -276.1608 305707.0
250 11740.3005 -276.1608

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
amber energies:
 NSTEP = 0 TIME(PS) = 0.000 TEMP(K) = 0.00 PRESS = 0.0
 Etot = -57686.8091 EKtot = 0.0000 EPtot = -57686.8091
 BOND = 341.5077 ANGLE = 716.0017 DIHED = 2372.7669
 1-4 NB = 1076.3715 1-4 EEL = 9100.8801 VDWAALS = 5527.3094
 EELEC = -76821.6464 EHBOND = 0.0000 RESTRAINT = 0.0000
 Ewald error estimate: 0.2976E-03

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
It's strange because in NAMD seems most important parameter I would play around with VDW energy is switching, and it's turned off. Same cutoff is set in two inputs, but still, I get very different results.
any suggestions would be appreciated! Thanks!
 

 

-- 
Miranda
      

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:51:14 CST