Re: NVT with truncated octahedron or with sphere?

From: Werner Crous (crous.werner_at_gmail.com)
Date: Wed Sep 16 2009 - 06:04:31 CDT

Dear Dimitar

In our group we have not used NAMD before, mainly CHARMM. CHARMM gives
the sides of the truncated octahedron in the section in the output
where it shows the unit cell information. I then set-up a cartesian
axes system where one of the vectors points in the direction of one of
X axis and the others make an angle of 109.47... with respect to each
other. One can easily derive the equations by using trigonometry. I do
not know how CHARMM points its vectors, but when I did a minimization
of 200 steps, then the whole truncated octahedron rotated and I could
run my dynamics. The clashing was therefore most likely because
CHARMM's vectors point in different directions. As we are more
familiar with CHARMM and its results in the group than NAMD (we have
only published papers using CHARMM) I wanted to use CHARMM as far as
possible. I however have a very large system and a proper simulation
with CHARMM takes too long. Therefore I am considering using NAMD. I
have found that with NPT in NAMD I get smaller volumes after 500 ps
than with CHARMM for the same system. Therefore I was reluctant to use
NAMD for NPT in a truncated octahedron. The discussion on the list I
am talking about is:
Re: inconsistence in NAMD and AMBER energy: AMBER truncted octahedron

Thank you
Werner

On Tue, Sep 15, 2009 at 7:34 PM, Dimitar V Pachov <dpachov_at_brandeis.edu> wrote:
> Dear Werner,
>
>
> ----- "Werner Crous" <crous.werner_at_gmail.com> wrote:
>
> | From: "Werner Crous" <crous.werner_at_gmail.com>
> | To: namd-l_at_ks.uiuc.edu
> | Sent: Tuesday, September 15, 2009 4:03:06 AM GMT -05:00 US/Canada Eastern
> | Subject: namd-l: NVT with truncated octahedron or with sphere?
> |
> | Dear all
> |
> | I am new to NAMD and want to use a truncated octahedron water box. I
> | equilibrated a 110A truncated octahedron box in CHARMM within the NPT
> | ensemble, then solvated my protein and ligand, ran NPT to adjust the
> | volume of the box and then NVT for 500 ps. I then tried to run NVT
> | with langevin in NAMD with triclinic unit cell imaging based on the
> | vectors in the CHARMM output file.
>
> I am not quite certain that I understand which vectors you used and where CHARMM outputs those vectors.
>
> I also do not quite follow your protocol of equilibration. I understand you created a box of water, equilibrated in NPT, then put protein and ligand in, equilibrated again in NPT, then fixed the volume in NVT. What was your final goal?
>
>
> | I used the following input for the
> | imaging and particle mesh ewald.
> |
> | set a 92.42967
> | set p 109.471220634
> | set pi 3.14159265358979323
> | set cosp [expr cos($p * $pi/180)]
> | set sinp [expr sin($p * $pi/180)]
> |
> | set Bx [expr $a*$cosp]
> | set By [expr $a*$sinp]
> | set Cx $Bx
> | set Cy [expr $a*$cosp*(1-$cosp)/$sinp]
> | set Cz [expr sqrt($a*$a - $Cx*$Cx - $Cy*$Cy)]
> |
> | cellBasisVector1 $a 0 0
> | cellBasisVector2 $Bx $By 0
> | cellBasisVector3 $Cx $Cy $Cz
> | cellOrigin 0.0 0.0 0.0
> |
> | PME             yes
> | PMEInterpOrder  6 ;#interpolation for B-spline
> | PMEGridsizeX 100 ;#should be a combination of multiples of 2 and 3
> | (and 5)
> | PMEGridsizeY 100
> | PMEGridsizeZ 100
> | PMETolerance 1.69e-11
> |
> | a is the side of the truncated octahedron as in my NPT simulation with
> | CHARMM.
>
> Which NPT simulation? The size changes during NPT, so which size did you use?
>
> | When I tried to run the simulation the energies were extremely high
> | and there were problems with the Settle algorithm. I then minimized
> | it
> | for 100 steps with conjugate gradient and then I was able to run
> | dynamics. Unfortunately, the kinetic energy was extremely high giving
> | temperatures of nearly 1000K and this for a system which was
> | equilibrated or near equilibration in CHARMM.
>
> How did you come up with the above cell vectors? If you do not match the CHARMM orientation of the truncated octahedron, then you will probably obtain clashes and high energies.
>
> |
> | Based on recent discussions on the list, it seems as if it is not
> | feasible to use truncated octahedron boxes in NAMD.
>
> Would you point to those discussions? What exactly do you mean by "it is not
>  feasible to use truncated octahedron boxes in NAMD."?
>
>
> | I now want to
> | know, that if I want to run a simulation in NAMD should I rather try
> | to use a very large sphere or what do you suggest based on
> | experience?
> | It seems as if the imaging might be the main problem. When I
> | visualized the box in vmd it looked as if it deformed significantly
> | at
> | some corners almost immediately.
>
> This indicates that the unit cell vectors you specified in NAMD do not form a truncated octahedron compatible with the CHARMM orientation and hence, over the course of the simulation NAMD will try to adjust the system based on the given unit vectors. Have you tried to continue the simulation for more than 100-200ps and see what happens? After certain time (~100ps-200ps), the system should stabilize and form a nice truncated octahedron. I personally have not tried or observed such a behavior, but people claim it happens.
>
> As I understand correctly, your idea is to build an equilibrated system in CHARMM and use the NAMD scalability for production runs, correct?
>
>
> |
> | Thank you in anticipation
> | Werner Crous
> | PhD student
>
> --
> ========================================================
> Dimitar V Pachov
>
> PhD Candidate
> Physics & Biochemistry Department     Phone: (781) 736-2326
> Brandeis University, MS 057           Email: dpachov_at_brandeis.edu
> ========================================================
>
>
>
>
>
>
>
>
>
>
>
>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:17 CST