Page 1 of 2 1 2 >
Topic Options
#8472 - 10/08/05 07:26 PM PBC for a linear protein
samcho Offline
Forum Member

Registered: 08/24/04
Posts: 6
Loc: San Diego, CA, USA
Hi all,

I am trying to simulate an ankyrin repeating protein via constant pressure PBC, and I am having difficulties setting up the simulation. This protein is several times as long as it is wide. I started off using the example script in the archive for PBC. The resulting solvent box is a rectangular cube. The simulation confines the center of mass of the protein to the center of the primary box. However, the protein rotates during the simulation until part of it escapes the primary box. Then a SHAKE error occurs and the simulation crashes.

I was wondering if anyone else had successes with explicit water simulations of a long protein. I would appreciate any advice for simulating such a protein.

Cheers,
Sam
_________________________
Sam Cho Department of Chemistry and Biochemistry University of California, San Diego

Top
#8473 - 10/09/05 08:35 AM Re: PBC for a linear protein [Re: samcho]
hkhandel Offline
Forum Member

Registered: 03/17/04
Posts: 98
You are probably doing image recentering of the protein by each residue (byres). You should do the image recentering of the entire protein by segment (byseg). If you do it by residue, as soon as part of the protein moves out of the box, CHARMM thinks that one atom of a bond is in one box, the other is in the other box, resulting in a bond length of ~ 60 Angstroms, hence the shake error.

Top
#8474 - 10/10/05 01:15 PM Re: PBC for a linear protein [Re: hkhandel]
samcho Offline
Forum Member

Registered: 08/24/04
Posts: 6
Loc: San Diego, CA, USA
Thank you for your reply. Changing the imaging specifications sounds like a very good idea to get rid of the SHAKE error. However, I would also like to keep the protein from rotating so that part of the protein lies outside the primary box. Is there a good method of doing this for a long protein?
_________________________
Sam Cho Department of Chemistry and Biochemistry University of California, San Diego

Top
#8475 - 10/10/05 01:52 PM Re: PBC for a linear protein [Re: samcho]
hkhandel Offline
Forum Member

Registered: 03/17/04
Posts: 98
Normally, one uses the NTRFRQ option (dynam.doc). But it is quite difficult to completely prevent rotation of the protein in free MD.

Top
#8476 - 10/10/05 02:15 PM Re: PBC for a linear protein [Re: samcho]
rmv Online   content

Forum Member

Registered: 09/17/03
Posts: 8392
Loc: 39 03 48 N, 77 06 54 W
Note that NTRFRQ applies to all atoms, i.e. the system as a whole, and not just the protein; the system (protein+solvent,etc.) cannot rotate, but the protein still can.

Depending on the properties you wish to calculate, the use of a restraint based approach such as CONS HARM ABSO (cons.doc) on selected atoms may or may not be appropriate; the impact of the restraints becomes an issue. If possible, it's better to use a near-sphere shape such as a rhombic dodecahedron (RHDO) or a truncated octahedron (OCTA) larger than the longest dimension. This way the protein cannot come into contact with an image copy of itself upon rotation. Note that in general some part of macromolecule may often protude from the unit cell; an image copy of the macromolecule will be poking into the cell from the opposing parallel face. This is expected, and accounted for in the calculation of energies and forces, as long as sufficient image transformations are defined. For CRYSTAL, the number of transformations available is controlled by the CUTOFF option used with CRYSTAL BUILD; I recommend at least half the length of your longest unit cell dimension. Not having enough transformations can lead to problems, including mystery SHAKE errors and other odd problems; shapes with all 90 deg corners (CUBIC, TETR, ORTH) require at least 26 transformations.
_________________________
Rick Venable
computational chemist


Top
#8477 - 10/10/05 03:30 PM Re: PBC for a linear protein [Re: rmv]
samcho Offline
Forum Member

Registered: 08/24/04
Posts: 6
Loc: San Diego, CA, USA
Dear Dr. Venable,

Thank you for your suggestions. I have never tried to simulate using PBC with rhombic dodecahedron (RHDO) or a truncated octahedron (OCTA). In both cases, however, I believe the the primary image shape is roughly a sphere. FYI, my protein is roughly three times as long as it is wide, so I believe a rectangular cubic image (if it does not rotate), might be the best approach.

I was thinking about your suggestion for a constraint. I would like minimal constraints so that I will not have to worry (much) about the influence of constraints. Your suggestion, CONS HARM ABSO for selected atoms (i.e. the end residues) might be the best approach because it will constrain the location of the residues in space. I will give it a try.

Sam
_________________________
Sam Cho Department of Chemistry and Biochemistry University of California, San Diego

Top
#8478 - 10/10/05 04:21 PM Re: PBC for a linear protein [Re: samcho]
rmv Online   content

Forum Member

Registered: 09/17/03
Posts: 8392
Loc: 39 03 48 N, 77 06 54 W
Note that the CONS HARM ABSO restraint force can be scaled so that flexing motions along the long axis could be permitted via e.g. ZSCALE 1.e-4 (assuming the long axis is along Z). Bending would be suppressed by the restraints in X and Y, however, so normal mode and other motional analysis results will be affected.

Note that with the anisotropic pressure scaling in CHARMM, the box lengths are free to change with the ORTH lattice type, and can do so with undesirable results at times. A single dimension can collapse such that a solute is in direct contact with image copies of itself along that dimension.

The RECT lattice type was added recently (c31b1), but hasn't been documented and the initial implementation had a bug; see my post in the "Bug Reports and Fixes" forum for the fix. It is identical to ORTH except that the ratio of cell lengths a/b, b/c, and a/c are constrained, and the pressure scaling becomes isotropic.
_________________________
Rick Venable
computational chemist


Top
#8479 - 10/11/05 05:17 PM Re: PBC for a linear protein [Re: rmv]
samcho Offline
Forum Member

Registered: 08/24/04
Posts: 6
Loc: San Diego, CA, USA
Dear Dr. Venable,

Using the ZSCALE keyword so that the end residues slide up and down the z-axis sounds like a great idea. I will look into the RECT lattice also. Thank you for your suggestions.

Sam
_________________________
Sam Cho Department of Chemistry and Biochemistry University of California, San Diego

Top
#8480 - 10/21/05 03:31 AM Re: PBC for a linear protein [Re: samcho]
samcho Offline
Forum Member

Registered: 08/24/04
Posts: 6
Loc: San Diego, CA, USA
Dear all,

I tried running a CPT simulation with PBC and the harmonic constraint described above. During the course of my simulation, the waters surrounding the protein diffused outside of the primary box, effectively evaporating. I had the same result when I used another harmonic constraint, but the simulation ran fine when there was no harmonic constraint. Am I missing something? Or perhaps a bug in the version I am using (c30a2)? I would appreciate any suggestions.

Sam
_________________________
Sam Cho Department of Chemistry and Biochemistry University of California, San Diego

Top
#8481 - 10/21/05 05:30 AM Re: PBC for a linear protein [Re: samcho]
lennart Online   content

Forum Member

Registered: 09/25/03
Posts: 4745
Loc: ~ 59N, 15E
Post your script.
_________________________
Lennart Nilsson
Karolinska Institutet
Stockholm, Sweden

Top
Page 1 of 2 1 2 >

Moderator:  John Legato, lennart