Improved statistical torsion angle potential yields higher-quality NMR protein structures

Wednesday, October 10, 2012 — Poster Session II
Noon – 2:00 p.m	Natcher Conference Center, Building 45	CIT	STRUCTBIO-2

Authors

• G A Bermejo
• G M Clore
• C D Schwieters

Abstract

Torsion-angle probability densities in a database of X-ray protein models can be converted into a so-called statistical potential energy terms that improve structural validation criteria and accuracy of calculated protein NMR structures. However, inspection of current energy surfaces reveals roughness and other features unsupported by angular populations in a modern database. Here, a complete reformulation of the potential addresses such deficiencies. Smooth, continuous densities, generated from a high-quality database via adaptive kernel density estimation, were converted into energy terms, robustly approximated by cubic interpolation during structure calculations. When tested on nine proteins, the resulting energy surfaces, which are both smoother and sharper, further improve conformational validation criteria, atomic interactions (clashes, packing), and agreement with cross-validated residual dipolar couplings.

back to top