[gmx-users] atomic charges

[Erik Lindahl]

T. Zaraiskaya wrote:

>hi all,
>i am using 43A2x parameter set for lipid topology;which method would be
>better to use for atomic charge calculations to keep charges consistent
>with gromos96 FF? Merz-Kollman, Mulliken,CHelpG....
>the geometry of the lipid was optimized with a rhf/6-31Gd basis
>set (gaussian98)
>
>tanya,
>
Well, the boring & difficult answer is "it depends".

On the positive side: When I did lipid simulations we tried a lot of 
different charges for DPPC (which is an uncharged lipid, though), and 
although the charges on individual atoms differ the overall dipoles and 
quadropoles are relatively similar. So, the most important thing is 
probably to find a set of charges that are consistent with each other if 
you want to compare different lipids.

A lot of the lipid simulations in the literature use charge sets like 
Chiu and Essex, which I believe are Mulliken-derived, but I also know 
that QM experts frown when they hear "Mulliken" today since it's really 
trivial to do better QM derivations...

If you have access to QM software and know how to use it I would 
probably be deriving my own charges (but I'm not a QM expert :-). If 
nothing else, it gives you the ability to easily derive new charges if 
you want to simulate a new lipid.

To do accurate lipid simulations it's important that you use long twin 
cut-offs (like 1.8/1.0), or better: PME.

Cheers,

Erik