The quality of homology models, especially ones with low sequence identity, are dependent on the quality of the alignment used. The models in MODBASE are calculated using alignments from PSI-Blast or IMPALA. Due to the heavy computational requirements of calculating models on a large-scale, as is the case with MODBASE, only one model is calculated per alignment.

Now, assuming that the alignment in MODBASE is very similar to yours, the best models you picked out of the 1000 are probably better than what is available in MODBASE.

Eswar.

On Jan 7, 2007, at 7:27 PM, hori koshii wrote:

> Hi all, > > Has anyone come across with the models on MODBASE? How does > that compare to our own models built by using "MODELLER"? (I know > they are both using MODELLER, but the protein I am working on, have > very different "insertion" loops from my results and modbase's) > > I have been trying to build a protein model using MODELLER8, but > recently come across with one already available on the MODBASE. I > think there is actually only one good template (over 30%) for my > target. I used MODELLER 8 to build 1000 models and pick out the > ones with lower DOPE scores. The loop conformations in my model and > the one on MODBASE are quite different. I am curious espeically > about the loop insertions. > So, I am wondering how come they are very different? Could anyone > from this forum give me any comment about this please? > > Thank you :) > > Sincerely yours, > __________________________________________________ > Do You Yahoo!? > Tired of spam? Yahoo! Mail has the best spam protection around > http://mail.yahoo.com > > _______________________________________________ > modeller_usage mailing list > modeller_usage@salilab.org > https://salilab.org/mailman/listinfo/modeller_usage