Statistical potential for assessment and prediction of protein structures

• DOI: 10.1110/ps.062416606
• Publication Date: 2006-10-31T00:27:49Z
• Authors: Min‐yi Shen; Andrej Sali
• Keywords: Decoy

Protein structures in the Data Bank provide a wealth of data about interactions that determine native states proteins. Using probability theory, we derive an atomic distance-dependent statistical potential from sample does not depend on any adjustable parameters (Discrete Optimized Energy, or DOPE). DOPE is based improved reference state corresponds to noninteracting atoms homogeneous sphere with radius dependent structure; it thus accounts for finite and spherical shape structures. The was extracted nonredundant set 1472 crystallographic We tested five other scoring functions by detection among six multiple target decoy sets, correlation between score model error, identification most accurate non-native structure set. For all best performing function terms criteria, except tie one criterion To facilitate its use various applications, such as assessment, loop modeling, fitting into cryo-electron microscopy mass density maps combined comparative protein incorporated modeling package MODELLER-8.