[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Problems optimizing when BLK residues are present

Cláudio Soares wrote:


>         We have been using MODELLER for quite a while (since version 4.0) and we
> have now encountered two problems that I would like to share with you.
> These problems are of different nature and the first one is the most
> important one. I am referring to problems using MODELLER version 6.0 under
> i386 Linux.

please visit modeller site and download the latest version, many bugs
were fixed.

> 1) We have proteins with non-standard co-factors, or simply co-factors that
> do not exist in the topology files (one example is FMN). When we derive the
> structure of the protein without the co-factors with standard optimisation
> procedures (such as refine 3), the structures come out all right with good
> Ramachandran plot and low values of the optimisation function. However,
> when we add these co-factors as BLK residues, the results are quite
> different, with much more Ramachandran violations and generally higher
> values of the objective function (which is not directly comparable between
> the two cases, but even so, the numbers are substantially different). This

they not comparable, indeed. Sometimes ligands introduce more
violations, that is not necessarily a reason to worry.  Try to run the
modeling several times from different starting point and select the
lowest energy model from this series of runs - consequently it  will
likely to have the least violated model. when modeling with ligands, you
disrupt a little bit the homology restraint network of the template,
which  was solved without a ligand. Still, you likely to end up with a
biologically more reasonable model, even though the model is somehow
more violated according to some homology restraints.

> is an observation from a set of runs and its seems statistically
> significant. Another supporting evidence for a less good optimisation in
> these cases is the observation that loop regions (not defined by
> homology-derived restraints) are much similar between different generated
> models when BLK residues are included than when they are not. Do you think
> that this is a consequence of using BLK residues or a bug in the code?
> 2) When adding our own restraints to the homology derived restraints, the
> program crashes with segmentation fault. An example follows:

please send a complete example: top, ali, pdb files 



Andras Fiser, PhD            # phone: (212) 327 7216
The Rockefeller University   # fax:   (212) 327 7540 
Box 270, 1230 York Avenue    # e-mail:
New York, NY 10021-6399, USA # http://salilab.org/~andras