model quality: GA341, DOPE and zDOPE
i'm trying to evaluate models generated by Modeller in comparison to some built using threading methods (ITasser). GA341 seems the primary coin-of-realm in Modeller-land, but depends on seqID and so this can't be used (i don't think?) on threaded models. DOPE can be, and normalized zDOPE seems recommended for comparison of models from different homologues.
but in trying to compare models, i'm finding Modeller models that have excellent (> 0.8) GA341 scores but pretty terrible (1 - 1.5) zDOPE scores. and so in comparison to ITasser models with good zDOPE scores (~ -0.5), these threaded models would seem preferred?
beyond the main CASP results, does anyone else have experience trying to compare threaded vs. homologue models in terms of these metrics, or others? thanks for any leads,
rik
On 03/29/2012 11:02 AM, richard k belew wrote: > i'm trying to evaluate models generated by Modeller in comparison > to some built using threading methods (ITasser). GA341 seems the > primary coin-of-realm in Modeller-land
Maybe 5 years ago... the newer DOPE method is superior. See Protein Sci 17, 1881-1893, 2008 (#205 at http://salilab.org/publications/)
> but in trying to compare models, i'm finding Modeller > models that have excellent (> 0.8) GA341 scores but > pretty terrible (1 - 1.5) zDOPE scores. and so in comparison > to ITasser models with good zDOPE scores (~ -0.5), these > threaded models would seem preferred?
I would trust zDOPE over GA341. But neither zDOPE nor GA341 cares whether the model is generated by Modeller or by some other method - they simply take PDBs as input.
Ben Webb, Modeller Caretaker
thanks for the quick response, Ben. i'll look up that pub.
> ... But neither zDOPE nor GA341 cares > whether the model is generated by Modeller or by some other method - > they simply take PDBs as input.
but GA341 does require the seqID line in the PDB file, which is not available for threaded models? but if GA341's been obsolesced by the new, improved DOPE i guess i don't even need to care!
thanks again for your great support of a fine platform.
rik
On 03/29/2012 12:36 PM, richard k belew wrote: > but GA341 does require the seqID line in the PDB file, which > is not available for threaded models?
That's correct, so you wouldn't be able to use it there anyway. But in principle you could feed it a PDB generated by a non-Modeller homology modeling method, for example.
Ben Webb, Modeller Caretaker
participants (2)
-
Modeller Caretaker -
richard k belew