Thanks Robin,
The heme system is a little more complicated because the minimized
structure of any perturbations of the ring will often result in changes
elsewhere.
I was hoping to use paramfit to work with this system and have had luck
getting results similar to previously developed heme parameters using it.
Though I have a fairly good number of optimized structures 200-500, I could
also use the optimization steps produced from Gaussian. This usually
requires a decent amount of manual effort to prune high energy structures,
but results in a ten to twenty-fold increase in available
structure-energies.
I am presently using a brute-force genetic-only approach to prevent
negative samples (I'd figured that new generations and mutations would be
reasonable). But there isn't a clear guide on what is a 'good' function
value.
The better results tend to have function values between 4-6 kcal/mol and
coefficients above 0.75. With simplex applied the results would be >1
kcal/mol with coefficients above 0.95, however as noted they were simply
impossible outcomes.
While it is (somewhat) clear when one result is better than another, it is
unclear what constitutes an acceptable result.
Also, would it be possible to include optional hard limits on bond
parameters for simplex optimization in future releases? It doesn't seem to
be a difficult problem at first glance).
-D
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Feb 12 2016 - 12:30:03 PST
