[CP2K:513] Re: Newbie Questions

[Jack Shultz]

Teo,

I appreciate your comments and I am in the process of learning about
molecular modeling. I am attempting to integrate these molecular modeling
tools into my distributed computing project http://hydrogenathome.org

I've hypothesized that it should be possible to identify catalysts based on
structural data. Specifically I believe that it is possible to model
transition state structures and their associated free energies and use that
data with semi-empircal molecualr docking to model the transition state
structures binding to reaction intermediates.

I know that due to the induced fit model of enzymes, the active site of an
enzyme goes through its own conformational changes to mold the substrate
into a transition state structure. Maybe the active site in its native state
is sufficiently close in structure to run a docking. Otherwise one would
also have to run some kind of sampling on the different conformational
changes in the protein structure.

I understand that I need to start small. I will probably look at the most
simple reaction with the most simple catalysts first. I may try a couple of
modeling theories
1) Parameterized DFT
2) semi-empirical QM/MM.

Perhaps treat it in a couple scenarios:
1) Run molecular docking on transition state structures with a catalysts
where the TSS model is derived in a reaction isolated from any catalysts.
2) Then another scenario, run a molecular docking between the substrate and
an enzyme. Generate a quantum box to of approximate 30 atoms and simulate a
reaction within that area, ignoring external influences

I don't know how realistic these ideas are, so really do appreciate having
holes poked into it and an exchange of ideas. Thanks for your help.

Jack



On Jan 7, 2008 1:21 AM, Teodoro Laino <teodor... at gmail.com> wrote:

>
> Hi Jack,
>
> On 7 Jan 2008, at 03:12, jackyg... at gmail.com wrote:
>
> >
> > I am looking at semi-empirical QM/MM as way to model catalyzed water
> > splitting hydrogen generating reactions. I've been veting this idea
> > among researchers and there seems to be problems with many of the
> > molecular modeling techniques such as QMC and DFT as applied to my
> > model.
>
> The information you are giving on the model are so few that i would say
> semi-empirical can be even worst in treating water hydrogen bonds
> compared
> to DFT. Semi-empirical would be definitely faster but don't expect a
> realistic
> representation of the HB pattern for your system (few people here
> have already
> tried QMMM semi-empirical and can comment on that). Moreover the bond-
> breaking
> energies.. I would be surprised to see numbers in better agreement
> than DFT.
> Of course if you go for your own semi-empirical model, fully
> reparametrized for
> your system the situation can be different.. but that's another story..
>
> > Here is more specifically what I am proposing. I hope to run this
> > analysis stochastically, perhaps treating the reaction in isolation,
> > going through different energy pathways, generating a different
> > transition state structures for each run along with associated free
> > energy profiles. Using the results of these reaction models, I will
> > run molecular docking analysis. So for a given reaction, we will have
> > generated a great quantity of data. We can thus compare the
> > theoretical vs the experimental, perhaps we can optimize these
> > simulations with heuristic algorithms.
>
> Since I don't have more details on your project I can just give
> general comments:
> In principle it should work..  BUT.. consider that the reaction on a
> smaller system
> (what you call reaction in isolation) is mainly driven by the
> potential energy (the entropic
> contribution can be very small). If you go on a larger system the
> entropic contribution
> can be pretty large and you have no information on that since there
> are no data you
> can extrapolate from your small cluster and your conclusions "could
> be" wrong.
> So be careful in making your final conclusion in the logic chain
> (models/extrapolations
> from cluster to larger systems).
>
> > I have just started installing the cp2k. If it is posible to build
> > this model with cp2k, I will really need help setting up these
> > simulations. I have not given up on DFT yet.
>
> Which model?
>
> cheers,
> Teo
>
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20080107/d246ff75/attachment.htm>