[gmx-users] Re: regarding DOPC lipid parameters

[Chris Neale]

First off: do you absolutely need dopc? If so, here is one way to 
generate the parameters.

1. Don't use popc.ito from Tieleman's website with a pdb from Feller's 
website. Tieleman PDB's and parameter files are based on 
(Berger_1997/Egberts_1994/Jorgensen_1988/Chu_1995/Chyu_1996) and those 
parameters are different from the Charmm parameters. Also I am sure that 
the atom names differ. So you would be best to either: a) Use all charmm 
stuff meaning you need to create the parameters in gromacs format (this 
is the more difficult option), or b) stick to what is on Tieleman's website.

2. Load one of the popc PDB files into your favourite viewer (I use VMD) 
and display only one lipid and then display all of the atom names. Then 
go through Tieleman's popc.itp and look at the [atoms] section and write 
down the atom number also beside each atom. Now copy this to a piece of 
paper (take up the full page) and photocopy it about 5 times.

3. Go through tieleman's popc.itp [bonds] section and mark each bond 
down beside the bond on one of the photocopies of your figure

4. Repeat step 3 on a different photocopy for each of [pairs] and 
[angles] and [dihedrals].

5. Now that you understand what each parameter is, it should be simple 
to go back and create your dopc.itp file. Don't forget to treat the 
double bond properly, especially note that it is not treated exactly as 
you might expect in the [pairs] section.

6. Next you need a pdb file. This again is not simple. I would 
personally recommend using pymol to build one lipid molecule and then 
build your membrane from scratch. However, you could also write a script 
to simply extend the tails in some already available popc.pdb membrane. 
In the second case you will still have problems with the rotamerization 
about the double bond. A 1ns equilibration at 310K followed by a high 
temperature pulse for 20ps at 510K and then cool to 360K over 100ps 
followed by 4ns at 360K and then a couple ns at 310K should get you to a 
proper distribution. This basic idea is from Takaoka et al Biophys J V79 
2000 3118-3138. I have tried it myself and it works (I used semiiso 
p_coupling) but obtaining a stable final state in which the membranes do 
not separate absolutely requires either (a) the initial 310K dynamics in 
order to get the tails to intercalate or (b) cool over 100ps, which is a 
longer cooling period than they used. I haven't found the need to 
determine which was the important part.

I realize that this procedure seems long and tedious. But my advice is 
that you cut corners here at your own peril.