namdEnergy gives significantly different values from those in log file?

From: Steven Samuel Plotkin (steve_at_physics.ubc.ca)
Date: Sat Nov 22 2008 - 23:39:40 CST

Dear NAMD users,

I have run namdEnergy on a dcd file, printing out All energies. First I
noticed the kinetic energy was not printed, then I compared the "TOTAL"
potential energy from namdEnergy with the TOTAL energy minus the KINETIC
energy in my log file. The two are quite different -- about 370 kcal/mol
on average. I'm confused by this, I'm not sure what is being calculated
differently? I've provided links to my dcd, log, namdEnergy settings,
and conf file are below. I've also posted a long version of my question
below to help put it in context.

(I'm using namd 2.6 and vmd 1.8.6).

Thanks for any help,

Steve P.

FILES:
log and dcd file:
http://www.physics.ubc.ca/~steve/NAMD/Repx/fold_dse2.job0.0.log
http://www.physics.ubc.ca/~steve/NAMD/Repx/fold_dse2.job0.0.dcd

Conf files:
http://www.physics.ubc.ca/~steve/NAMD/Repx/fold_dse2.conf
http://www.physics.ubc.ca/~steve/NAMD/Repx/dse2_base.namd

Screenshot of my namdEnergy gui, with settings:
http://www.physics.ubc.ca/~steve/NAMD/Repx/Screenshot-NAMDEnergy.png

xsc file (if needed):
http://www.physics.ubc.ca/~steve/NAMD/Repx/fold_dse2.job0.0.xsc

/Long version of my question:/

I have been able to modify the Alanin replica exchange tcl code to run
on a small peptide which has been solvated and is in ionic solution. I
am using namd 2.6 on an x86_64 amd cluster. I ran 16 replicas on 8 CPUs
across 2 nodes of a cluster, with temperatures ranging from 280K to
440K. My conf file has some new information not in the alanin example
because of my solvated box: I have cellBasisVectors, wrapWater, wrapAll,
PME, and Pressure control options including langevinPiston. I was
careful to choose the langevinPistonTemp to be $NEWTEMP in
replica_exchange.tcl.

I want to construct a free energy profile for the small peptide. However
this must be obtained at one (target) temperature, so I reweighted all
states by what their Boltzmann weights should have been at the target
temperature. This doesn't work for me: because the system includes the
peptide plus the whole water box, after re-weighting, one particular
state ends up having all the Boltzmann weight in the partition function.
I didn't think this was feasible.

So then what I must use is the total energy (KE +PE) of the protein
itself, which includes self interactions /plus/ interactions with water
and ions. Then I can reweight these energies at the target temperature
and get a new partition function and free energy profile. I thought
namdEnergy would be the tool for this.

As a sanity check, I ran namdEnergy on one of my dcd files, printing out
'All' energies. First I noticed the kinetic energy was not printed, then
I compared the "TOTAL" potential energy with the TOTAL energy minus the
KINETIC energy in my log file. The two are quite different -- about 370
kcal/mol on average. However the average total potential energy is
about 4460 kcal/mol, so the mean error is about 8%. Am I doing something
wrong here? Or is this to be expected?

At the end of the day what I believe I want is simply the namdEnergy
numbers for selection 1= PROTEIN , selection 2= "everything else",
added to the PROTEIN-PROTEIN namdEnergies. However the errors present in
my sanity check are giving me pause that the namdEnergies are trustable.

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:48:40 CST