On 13 Jul 1999 03:42:49 -0700 "M.A.Hughes" <m.a.hughes at bangor.ac.uk> wrote:
> Hello,
>> Does anyone have any experience of modelling small (2-6 residue)
> peptides using random conformational analysis procedures? I have
> modelled several tripeptides using the Sybyl package and their Random
> Search procedure (1000-5000 search iterations) with a distance-based
> dielectric constant of 80 to simulate water. I have now submitted a
> dipeptide to a similar conformational search except that I have included
> explicit waters (using the SILVERWARE algorithm and default settings)
> and a constant dielectric function of 80 to simulate bulk water outside
> of this solvation box. Specifically, I am interested in knowing two
> things:
>> 1. How much longer will this sort of search take (it's been running
> nearly a week now!)?
> 2. Are the conformers found likely to be very different from those that
> would have been found using the simpler approximation to water of a
> distance-based dielectric function? Since I am interested in modelling
> a number of tri- to hexapeptides, the trade-off between computational
> time and "reliabilty/accuracy" of results is paramount.
>> Thanks very much for any help,
>> Neil
> ------------------------------------------------------------------------------
> Neil J. Marshall
>> Memorial Building Tel: +44 (01248) 351151 Ext. 2367
> School of Biological Sciences Fax: +44 (01248) 370731
> University of Wales, Bangor E-mail: bss219 at bangor.ac.uk> Gwynedd, LL57 2UW
> United Kingdom
> ------------------------------------------------------------------------------
Hi,
It's some time since I dabbled with SYBYL but I have the following comments.
a) Absolute calculation time will depend on the hardware which you haven't
mentioned. However energy calculations on an assembly of n atoms involve approx. n
x n interactions. If solvation increases the number of atoms to say 2n, calculation
time will therefore take about 4 times longer!
b) Realistic modelling of the dielectric constant is not trivial. Certainly you
should never use a distance based constant of 80!!! The distance based constant
(n/r) tries to model the bulk value (80 for water) at large values of r while
allowing for increased interactions between charges which are close together and
not separated by bulk water. By setting n = 80 you are grossly underestimating
electrostatic interactions (almost completely neglecting them). I admit the manuals
do not give explicit guidance. I tried modelling a small peptide containing Asp and
Lys using n = 1 and found the sidechains of these residues were glued together,
which certainly wasn't reflected in the NMR data. I increased n until such
"unreasonable" behaviour ceased to be observed and found n= 4 prevented the
electrostatic term dominating the interactions. I think I have seen n = 4 and n = 8
used in literature.
c) I'm not sure it is sensible to do a conformational search with explicit
solvation, since a conformational search is designed to find low energy
conformations based on intra-molecular interactions. Solvation could of course
alter the relative energies of conformations found by the search. The usual outcome
of a conformational search is a series of low energy conformations (or of clusters
of related conformations). I think it would be much quicker (and just as valid) to
do the search without explicit solvent and then solvate and minimise each of the
low enenrgy conformations found, to obtain realistic energies for comparison.
Regards
David
----------------------
David J.S. Guthrie,
The Queen's University of Belfast,
Centre for Peptide and Protein Engineering,
School of Biology & Biochemistry,
97 Lisburn Road,
Belfast BT9 7BL, N. Ireland, U.K.
Tel (0)1232 272043
Fax (0)1232 236505
e.mail d.guthrie at qub.ac.uk
---
