[gmx-users] forcefield validation

[Mark Abraham]

Qin Shanshan wrote:
> Thanks very much for your patient explaination.It really is a question 
> confusing me much. What I am interested in is inserting some organic 
> molecules into lipid bilayers. Tieleman's lipid parameters have given 
> out the interactive LJ parameters with ffgmx forcefield. So if I want to 
> simulate lipid with tileman's force field I should use ffgmx for organic 
> molecules. However I have to prove the ffgmx forcefield for organic 
> molecules is correct too. Somebody prove this by calculating the new 
> molecules' melting point or their structures in  solutions to compare 
> with X-ray results. I found
> Monika Holtje's BBA article(Biochimica et Biophysica Acta 1511 (2001) 
> 156^167), they simulated cholesterol and stearic acid and palmitic acid, 
> constructing topology files using Gromos 87 forcefield , and put the 
> Cholesterol topology on Gromacs website.They said " The geometries, 
> conformations and configurations as they occur instandard GROMACS 
> dynamics in vacuo and in water using the parameter set in this file, 
> compared well with crystal structures". In their topology , they set the 
> partial charge like this:
>  
>  
> ;   nr    type   resnr  residu    atom    cgnr    charge   ; total charge
>      1     CH3       1    CHOL     C1        1     0
>      2      CB       1    CHOL     C2        2     0
>      3     CH2       1    CHOL     C3        3     0
>      4     CH2       1    CHOL     C4        4     0
>      5     CH1       1    CHOL     C5        5     0.14 ; charges adopted
>      6      OA       1    CHOL     O6        5    -0.54 ; from Retinol in
>      7      HO       1    CHOL     H         5     0.40 ; ffgmx.rtp
>      8     CH2       1    CHOL     C8        6     0
>      9      CB       1    CHOL     C9        7     0
>     10    CR61       1    CHOL     C10       8     0
>     11     CH2       1    CHOL     C11       9     0
>     12     CH1       1    CHOL     C12      10     0
>     13     CH1       1    CHOL     C13      11     0
>     14     CH2       1    CHOL     C14      12     0
>     15     CH2       1    CHOL     C15      13     0
>     16      CB       1    CHOL     C16      14     0
>     17     CH3       1    CHOL     C17      15     0
>     18     CH1       1    CHOL     C18      16     0
>     19     CH2       1    CHOL     C19      17     0
>     20     CH2       1    CHOL     C20      18     0
>     21     CH1       1    CHOL     C21      19     0
>     22     CH1       1    CHOL     C22      20     0
>     23     CH3       1    CHOL     C23      21     0
>     24     CH2       1    CHOL     C24      22     0
>     25     CH2       1    CHOL     C25      23     0
>     26     CH2       1    CHOL     C26      24     0
>     27     CH1       1    CHOL     C27      25     0
>     28     CH3       1    CHOL     C28      26     0
>     29     CH3       1    CHOL     C29      27     0

This charge distribution looks particularly weird and/or indefensible.

> However when I use B3lyp/6-31G(d,p) to calculate mulliken 
> distribution 

...which doesn't have any physical meaning...

> of cholesterol the charge distribution is rather different. 
> How did they decide their partial charge? 

You'd have to ask them, or read their paper.

> Is it customed to let alkane 
> groups zero charged?

No.

> Many articles didn't give out the validation 
> procedure of their topologies in detail, are there some classic articles 
> explaining how the forcefield we used today, such as OPLS, GROMOS, AMBER 
> and CHARMM , developed? I want to know how they were deduced from the 
> very beginning.

They're all referenced in the GROMACS manual, like you'd expect.

Mark