Date: Wed, 31 Jul 2002 08:26:49 -0600 (Mountain Daylight Time)
From: "Thomas E. Cheatham, III"
Subject: Re: periodic box

Date: Tue, 30 Jul 2002 22:18:41 -0700 (PDT)
From: Thomas Cheatham
To: Jiyoung Heo
Subject: Re: periodic box

> I am trying to do simulation with DNA dodecamer crystal structure. I
> want to use the periodic box which is given in the pdb file of this
> crystal. The problem is that the bounding box size of the model in one

Note that crystals in PDB files may not contain the complete (symmetry
related) set of molecules that fill the unit cell, i.e. sometimes it is
necessary to apply symmetry and build up the full cell; assuming you do
this (or you downloaded a complete unit cell from the nucleic acid
database), then various means can be applied to build the complete

It is a tricky process as it is not clear at the outset how much
additional water (or ions) should be added. A discussion of some of the
issues is presented in a paper by Bevan/Darden and co-workers in Biophys.
J. 78, 668 (2000). In that paper, a careful equilibration protocol was
applied where constant pressure MD was applied to relax the initial
solvent/ion density and then water added/removed to reach a reasonable
density and match the crystal unit cell dimensions.

If you have too much water, you will greatly inhibit fluctuations; if too
little, vacuum bubbles could crash into your structure and destroy it...

> direction is larger than the size of the crystal cell.
> dimensions of the model: 32.763400 33.814293 49.569000
> (I have got from 'setbox my_model' command)
> dimensions of crystal cell: 25.300 41.700 66.000

I do not understand the above. Why is your model "bigger"? Did you add
additional water? Or is the initial size larger (perhaps due to symmetry
in the DNA that isn't being enforced in your crystal, i.e. if the duplex
is symmetric, the unit cell may be P2 and represent only half the DNA?) or
is the alignment different in your PDB model than what AMBER expects?

I will assume that you added extra solvent to "build" it to some size.
Then it may be a question of alignment in that your DNA may not be in the
same orientation in AMBER as the crystal expects. Make sure that the
orientations are equivalent.

> I tried the 'setbox' command with specified buff values. Of course, in
> the x direction, the buffer value is negative.
> However, the box size did not change and it is still 32.763400 x
> 33.814293 x 49.569000.
> Is there any way to change the size of the box as what I want?
> Some program I am using recognizes the 'CRYST1' line in the pdb file and
> builds a crystal cell automatically. Does the amber have this option?

AMBER does not have automatic options to understand crystal symmetry. One
has a little more control in CHARMM, but even in CHARMM is it not
straightforward to build a crystal.

I would suggest the following:

(1) make sure you have a complete unit cell (or cells)
(2) make sure the alignment is as expected (before adding additional water)
(3) add water using the solvateBox with separate x,y,z directions,
changing each x,y and z until the created box has the size you expect
(i.e. run LEaP repeatedly with different x, y and z's until you find a set
that builds the box size you'd like)
(4) carefully equilibrate (with DNA restrained or fixed) to relax water
and then add/remove waters to get the size you want.
(5) think about what ionic conditions are present and whether you need to
add salt...

If you have questions on any aspect, let me know. Note that there is an
undocumented feature of rdparm called "testwater" that will write an old
style prmtop with a chosen number of waters based on a solvated prmtop.
This can be used to create a prmtop with a particular number of waters.
Then a PDB file can be edited (to remove/add waters) to match and then
ptraj used to create a restrt.