Page 1 of 1


PostPosted: Sun Oct 26, 2008 8:38 pm
by kevin_karplus
I was rather hoping that the AL2TS conversion would be fixed this year to produce complete (if poor) models rather than backbone-only models.
Our model-quality assessment has difficulty assessing the quality of backbone-only models (it needs at least CB atoms).

No one really does backbone-only modeling from alignments anyway---the standard protocol is to use something like SCWRL to put on the sidechains (usually copying sidechains that are unchanged and just doing scwrl optimization for the sidechains that are changed). That would provide fair models for alignment-only servers which could be properly assessed in the MQA tests.

I've noticed also this summer that AL2TS occasionally produces models that are then rejected by the next part of the model-checking pipeline, as there are differences between AL2TS and the next program in the pipeline about whether to include residues which don''t have all 4 backbone atoms.