by kevin_karplus on Sun Sep 07, 2008 10:52 am
I like Zhang's assessment by TM_score and HB_score (perhaps because it puts my server second, right behind Zhang's). It seems that Zhang fixed the problem in CASP7 of bad models built on good CA traces, if he is now doing best at the Hbonds.
I think HB_score should be used only on prediction of a "new category", say "H-bond prediction" , just like side-chain modeling, not on
traditional 3D-structure prediction which most focused on.
The CASP7 assessors ranked groups by HB and GDT. It's time for CASPs to set up a somewhat consistent criterion.
I disagree—I think we need new assessment methods that better distinguish good models from great models. GDT is a fine measure for the template-free models and for models that are not so great, but once the models start getting good (GDT > 85%, say) then ranking just based on the CA trace is sort of stupid. Getting the model right in traditional 3D modeling is the goal, and getting it right is not just getting the CA atoms in roughly the right places.
Correctness of hydrogen bonds is one measure that helps distinguish among good models. Other measures (all-atom RMSD, chi1 correctness, ... ) can also be applied. My former student, Firas Khatib, has come up with some new topological measures (based on slip knots) that are almost completely orthogonal to the GDT measure, but which usually distinguish experimental models from CASP models. They measure a property that none of us are getting right yet, but which is invisible to GDT. (Sorry, he hasn't made a program that can be used by any one but him yet—put pressure on David Baker, who has hired Firas as a postdoc.)