[gmx-users] charges in .gro

[Erik Lindahl]

Hi Daan,


On Dec 8, 2005, at 9:23 PM, Daan van Aalten wrote:

>
> Hi Rama
>
> I would be grateful if you could back up that "not reliable" statement
> with some actual examples.

This is certainly not an argument in the current discussion, but  
reading it I remembered that I had problems with a carbonate ion in  
Lactoferrin a while ago. This are the lines:

ATOM   5317  C   CO3   695     -29.141  24.537   0.765  1.00  
29.28      1LCF5590
ATOM   5318  O1  CO3   695     -27.891  24.233   0.580  1.00  
30.58      1LCF5591
ATOM   5319  O2  CO3   695     -29.845  23.937   1.726  1.00  
34.74      1LCF5592
ATOM   5320  O3  CO3   695     -29.904  25.364   0.095  1.00  
26.35      1LCF5593

If I recall correctly I ended up with neutral charge by default, and  
charge -1 (instead of -2) when using full charges.

Now, I _do_ know I should just have used a proper carbonate ion in  
the G96 FF, but I was a bit lazy and just submitted all  
heterocompounds...


As for the "reliability": you can no doubt get pretty good charges  
from QM if you start with proper Lewis electron diagrams, minimize  
all possible conformations, spend a week doing gaussian calculations,  
ponder whether you should use implicit QM solvation potentials,  
extract potentials on a dense grid, and finally do two-step RESP- 
fitting. This far I've spent >3 year of CPU time just for DPPC  
headgroup QM runs, and we have just begun fitting the torsion  
potentials.  On the other hand, ProDRG provides a very good  
approximation for most molecules in ~3 seconds.

Life is all about choices!

Cheers,

Erik



-----------------------------------------------------------
Erik Lindahl  <lindahl at sbc.su.se>
Assistant Professor, Stockholm Bioinformatics Center
Stockholm University, SE 106 91 Stockholm
Phone: +46 8 5537 8564     Fax: +46 8 5537 8214