[gmx-users] difficulty to reproduce a box described in the literature

Berk Hess gmx3 at hotmail.com
Thu Apr 15 12:08:01 CEST 2004

>I would like to run the same MD simulation described in Van der Spoel et
>al 1996 Protein Science, 5: 2044-2053.
>In the methods it is said that "the helix was solvated in a cubic box
>with an edge of ~ 5.0 nm filled with 4012 SPC water molecules"
>If I understand correct I have to specify using editconf -bt cubic and
>-d 2.5. Correct or not?

No, not correct.
As you can see from the text below and with editconf -h, option -d
sets the distance from the solute to the edge of the box.
Therefore you get a box size of: 4.449+2*2.5=9.449.
This explains the huge number of water molecules.
The correct command is:
editconf -box 5

>I read that 2.5 is the distance of the protein to the edge of the box,
>in fact when I run editconf it says
>"diameter    :  4.449  (nm)"
>But then when I run genbox -cs spc216.gro it puts 26825 water molecules
>So how David obtain only 4012 SPC water molecules?

A cubic box with edges of 5 nm is very small for a protein with diameter 
4..449 nm.
You would have to orient it along a box diagonal. But the chance is high 
that it
will rotate within a few ns and the ends of the helix might have a strong 
through the periodic boundary conditions.
A much better setup, which was not possible when Van der Spoel did his 
is to use a rhombic dodecahedron with a larger box vector length.
A rhombic dodecahedron with a box vector length of 5.6 has the same number 
water molecules at a cubic box of 5 nm. But because computers and Gromacs
have become much faster one should use a rhombic dodecahedron of at least 6 


Hotmail en Messenger on the move 

More information about the gromacs.org_gmx-users mailing list