Dear Crystallographers,
I have a case in which a lower sequence ID model worked better than >100 different higher-seqID models in MR searches. The model that worked was 88% identical, whereas the non-working ones were 98-100% identical (it's a highly flexible protein). But it gets even more strange: the structure is now solved, and the "88%" model is not at all the most structurally similar model in the pdb, and yet some much more structurally-similar models failed. I am considering the hypothesis that the use of seqID as a proxy for structural similarity by MR software may not work so well when the sequence ID is really high (98-100%), as in the case of the failed models. This may be further complicated by the presence of pretty heavy twinning (~45%). I am curious whether anyone knows:
-What is the function by which sequence identity is converted to structural similarity in MR software?
-If this function somehow under- or overestimated the structural similarity, could that lead MR software to fail, perhaps by giving inappropriate priors?
-Would the conversion of seqID to rmsd need to be tweaked for twinned data, perhaps?
-Any other ideas how this could happen?
All the best,
Jacob Keller
*******************************************
Jacob Pearson Keller, PhD
Looger Lab/HHMI Janelia Research Campus
19700 Helix Dr, Ashburn, VA 20147
email: [log in to unmask]
*******************************************
|