[gmx-users] PRODRG and partial charges

[Douglas Ridgway]

I'm confused about how to assign partial charges. I did some browsing of 
the archives, so I know this has come up before, and have started to read 
some papers, but I'm still confused.

One suggestion is to use PRODRG, another is to get partial charges from
QM, and I've also read that QM charges are too big and should *not* be
used. PRODRG offers both "full" and "partial"  charges, though I'm not
sure what these mean exactly.

My interest is drug molecules in membranes. At the moment, though, I'm 
playing with octanol. Octanol has some charge at the end, due to the 
hydroxyl. The question is how much. Here's a table of some possible 
charges for the last carbon, the oxygen, and its attached hydrogen:


             C        O      H
partial    0.026  -0.091  0.013
full       0.043  -0.151  0.022
a          0.040  -0.060  0.020
G45a3      0.150  -0.548  0.398
G45a3MOD   0.156  -0.563  0.407
I2         0.117  -0.323  0.181

"partial" and "full" came from PRODRG today. "a" also came from PRODRG, 
but a couple weeks ago, and I can't seem to reproduce it. (My notes show 
I've gotten other results on full charges too. Perhaps I'm doing something 
different or wrong?) 

The lines G45 etc. come from McCallum and Tieleman, JACS 124 15085-. The 
G45a3 line is also the same as the example given in the PRODRG paper of a 
charge group. The line I2 was generated by a collegue using Insight II to 
run some quantum calculation in a way I don't entirely understand.

The spread in values looks fairly wide, with the PRODRG values much lower
than the McCallum paper, and the QM values inbetween. It also seems to be
significant: I've run short simulations of octanol and water using both
"a" and "G45a3MOD". The results are visually quite different: "a" looks
fairly disordered, while "G45a3MOD" octanols line up straight and
parallel, at an angle to the octanol water interface. ffgmx force field,
SPC water, PME, rvdw=1.2.

Do charges need to be parameterized with the rest of a force field?  Will
they depend on cutoffs, other molecules and such? How do you do this for a
new molecule? How arbitrary is it? And how do you figure out what results
to believe?

Sorry for all the questions! I'm still trying to get a clue about MD...

doug.