[gmx-developers] genvsites

[David van der Spoel]

Berk Hess wrote:
> Erik Lindahl wrote:
> 
>> Hi Erik,
>>
>> David and I spoke about this last friday, and I'm not sure I agree 
>> completely :-)
>>
>> First, this approach is just moving the problem. Instead of having to 
>> create dummy settings for each force field you end up having to 
>> specify dummy parameters for each new molecule/residue. This will 
>> really be a pain for people working with small organic compounds, etc.
>>
>> The current aromatic vsites are horrible (construction-wise, I'm not 
>> even thinking about their usefullness...) and should be replaced, but 
>> the "standard" hydrogen vsites can actually be constructed completely 
>> automatic for new molecules.
>>
>> I've looked at quite a few vsites manually the last couple of days, 
>> and the largest deviations by far occur in peptide groups and similar 
>> constructs that aren't exactly planar in reality, but the vsite 
>> construction forces them to be. There's nothing we can do about that, 
>> and in comparison the deviations between ideal and force-field values 
>> e.g. for CH2 angles are very small.
>>
>> The other problem concerns e.g. NH2 groups. If I recall correctly 
>> these are sometimes planar in Gromos96, while OPLS/Amber model them as 
>> non-planar. In my opinion the right thing to do in that case is to 
>> model the force field as good as possible, since the point is to 
>> remove degrees of freedom while altering as little as possible. 
>> Otherwise we should for consistency also alter the constraint lengths 
>> of all bonds to agree with experimental structures rather than the 
>> equilibrium values in the force field.
>>
> I can remember a discussion on this, but I think NH2 are should be and 
> are always planar.

In a de-protonated Lys they are not. They are planar only in resonant 
systems like Asn/Gln/Arg side chains. In theory we therefore need to 
support both.


>>
>> So, it's definitely worth cleaning up, it would be nice to generate 
>> the MC/MN dummy mass parameters automatically from the force field 
>> (easy), and then create a residue-specific database for "advanced" 
>> constructs such as virtual aromatic sidechains. I still think it's 
>> more correct to set the simple vsites from the force field, though, or 
>> possibly the input structure.
>>
> I agree.
> 
> I don't know about the usefulness of vsite aromatics.
> I have never used them and will probably never use them.
> So as far as I am concerned they could be removed,
> but maybe some people do use them.
> 

So it seems that the common denominator is that things should be simple 
and general. But it doesn't hurt if one can do more complex descriptions 
as well (if only to test the generic ones).

In addition if the input structure is correct one can deduce the 
chemistry from that without too much trouble (some of that functionality 
is in x2top).

For existing FF files it might be slightly better to use the FF 
information, although this will require a tpr file for input.

-- 
David van der Spoel, Ph.D.
Molec. Biophys. group, Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205. Fax: +4618511755.
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://folding.bmc.uu.se