Increased Reliability of Nuclear Magnetic Resonance Protein Structures by Consensus Structure Bundles

Nuclear magnetic resonance (NMR) structures are represented by bundles of conformers calculated from different randomized initial structures using identical experimental input data. The spread among these conformers indicates the precision of the atomic coordinates. However, there is as yet no reliable measure of structural accuracy, i.e., how close NMR conformers are to the “true” structure. Instead, the precision of structure bundles is widely (mis)interpreted as a measure of structural quality. Attempts to increase precision often overestimate accuracy by tight bundles of high precision but much lower accuracy. To overcome this problem, we introduce a protocol for NMR structure determination with the software package CYANA, which produces, like the traditional method, bundles of conformers in agreement with a common set of conformational restraints but with a realistic precision that is, throughout a variety of proteins and NMR data sets, a much better estimate of structural accuracy than the precision of conventional structure bundles. [Display omitted] •Protocol for NMR structure determination with automated NOESY assignment•Consensus structure bundle whose radius is a good measure of structural accuracy•Accuracy is not overestimated by unreasonably high bundle precision•The protocol was applied to 4,066 data sets of 18 different proteins Overestimation of structural accuracy by unrealistically tight structure bundles is a common issue with available methods for NMR structure calculation. Buchner and Güntert develop an approach that minimizes this problem by using consensus restraints from multiple automated NOESY assignment/structure calculation runs to recalculate a consensus structure bundle.