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

model evaluation

Dear Modeller users,

I have generated some models using Modeller, and I wonder if you could
recommend the best way to evaluate them.  My sequence homology between
modeled protein and the templates is about 30%, so the confidence in some
regions of the sequence alignments is low.

I have generated several plots for energy versus residue number using
ProsaII, and these show quite nicely which regions have poorer structure.
These correlate to regions of low sequence conservation in the alignments.

My question to you is how valid the more specific analyses of atomic-level
contacts in the models would be.  I have run Procheck and Whatif analyses
as well, and the quality check scores are within acceptable ranges.  In
cases with 30% homology I would expect many errors in atom placements, so
should I even consider these stereochemical and packing constraints?

Can you recommend a good reference that discusses these issues?

Thanks very much for your help.


Giselle Knudsen
graduate student, CCB Program
University of California San Francisco