On 11/01/2011 10:28 AM, Andrew Voronkov wrote: > how on your opinion modeller objective function correlates with > quality of the models. For example I have objective function value > 1730 for one modelk and -750 which as I understand reflects free > energy. Does it mean that I shouldn t consider the first one at > all?
The Modeller objective function is simply a measure of how many restraints were violated in building your model. While the score could be thought of roughly as an "energy" (not free energy) this is certainly not a rigorous measure of model quality. Use an assessment function such as normalized DOPE for this purpose.
> Let's say I want to make cutoff -100 for the objective function. In > which format it should be done? > > automodel.max_molpdf -100
Build your models, then assess them and pick the best ones. max_molpdf is designed as a filter to throw away the very worst models - it doesn't act on the final PDF, but the score of intermediate (partially optimized) models - the reasoning being to not waste time on trying to optimize a model that is very bad. Thus max_molpdf is not really what you want here.
Ben Webb, Modeller Caretaker