I stumbled upon what I think is an interesting bug in Amber 8. I used tLEaP
to set up a system consisting of a protein solvated in TIP3P water, using
ff03. The PDB that I used still had its crystallographic waters, and I used
the solvateoct command to add a waterbox. I did not change the PBradii from
its default value.
In the prmtop file that tLEaP wrote out, the %FLAG RADII section showed that
for the crystallographic waters, the radii for the hydrogens were 0.8 and
1.2. They had different radii. Most of the waters (added by solvateoct)
had the correct radii of 0.8 for both hydrogens. Clearly, something weird
is going on here. Try it yourself by making a PDB file with just one line:
ATOM 1 O WAT A 1 0.000 0.000 0.000 1.00 0.00
Load it into LEaP and write out a prmtop file.
According to the default radii definitions (mentioned in section 3.6.45 of
the Amber manual), hydrogens that are bonded to O are given radii of 0.8,
but hydrogens bonded to other atoms are given other radii. The reason that
the second hydrogen in water is given a different radius is that the
definition of TIP3P water in the solvents.lib file has the connectivity
section
!entry.TP3.unit.connectivity table int atom1x int atom2x int flags
2 1 1
2 3 1
3 1 1
The first mention of atom H3 (the second H) has it single-bonded to H2, and
only afterwards lists the single bond between H3 and O1. Consequently, the
hydrogen H3 here is assigned a radius as if it were bonded to H2, not O1.
In tip3pbox.off, the connectivity for the atoms is in a different order:
!entry.TIP3PBOX.unit.connectivity table int atom1x int atom2x int flags
1 2 1
1 3 1
2 3 1
Here, the first mention of H3 is its bond to O1, which is why tLEaP assigned
the appropriate radii to all the hydrogens that were added with the
solvateoct command.
Changing the order of the connectivity in solvents.lib was sufficient to
eliminate the bug.
The radius section of the prmtop file isn't important to my research, but it
would probably be safer to change the order of the atom connectivity of
water definitions (putting the H-O bonds first) in future versions of Amber.
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Jan 17 2007 - 06:08:30 PST
