For NMR models

Since NMR does not produce a density map, validation relies heavily on how well the atomic model(s) satisfy the experimental restraints derived from the NMR data.

Assignment completeness

For NMR structures, assignment completeness is a crucial metric that reflects how much of the expected chemical shift data for a molecule has actually been assigned to specific atoms in the model. Chemical shifts are the primary experimental data in NMR, providing information about the local electronic environment of atomic nuclei (e.g., hydrogen, carbon, nitrogen).

A higher percentage of assigned chemical shifts (closer to 100%) generally indicates a more complete model. Lower completeness might mean that the experimental data collection or analysis was incomplete. However, completeness alone does not guarantee a well-defined 3D structure. The final model’s quality also depends critically on the number and distribution of structural restraints. For instance, flexible regions may have complete assignments but remain structurally ambiguous due to a lack of sufficient restraints.

Restraint violations

This is arguably the most crucial metric for NMR structures, as it directly assesses how well the calculated 3D model(s) adhere to the experimental data used to build them. It quantifies how many of the experimentally derived restraints (e.g., distance restraints from NOEs – Nuclear Overhauser Effects, or dihedral angle restraints) are violated by the calculated structure(s).

NMR restraints are spatial constraints derived from experimental data, such as NOE distance measurements and dihedral angle measurements derived from chemical shifts. These restraints act as guidelines for the structure calculation software, directing it to find a structure that satisfies the experimental data.

Violations occur when a structural parameter (e.g., a distance between two atoms) in a model structure falls outside the allowed range defined by a restraint. For example, if an NOE restraint specifies that two atoms should be within 5 Ångstroms of each other, and in the calculated structure, they are 6 Ångstroms apart, a violation occurs. Violations can be categorised as either upper-bound or lower-bound, depending on whether the measured value exceeds the upper limit or falls below the lower limit of the restraint. A lower number and magnitude of restraint violations indicate a more accurate and reliable NMR model that is consistent with the experimental data. Large or numerous violations suggest issues with the model building or the interpretation of the NMR data.

Restraints that are violated in at least one model are counted as violated, and restraints that are violated in all the models are counted as consistently violated.

Distance violations

Distance violations occur when the distance between two atoms in the model is significantly different from the distance range determined experimentally by NMR. These are typically reported with a threshold (e.g., a violation if the distance differs by more than 0.5 Å). If the measured distance in a model lies outside the boundaries defined by the restraint, then the absolute difference between the measured value and the nearest boundary is reported as the violation value.

The results are reported and binned into small, medium, and large violation categories based on the magnitude of the violation values. In each bin, the average number of violations per model is calculated by dividing the total number of violations in each bin by the size of the ensemble. The maximum value of the violation in each bin is also reported.

The table below lists distance violations per bin in the de novo designed protein PDB ID 7M5T.

Dihedral-angle violations

Dihedral angle violations occur when a bond rotation angle (Ramachandran ϕ or ψ angle) in the model deviates significantly from an experimentally determined range. For proteins, dihedral-angle restraints can be derived using backbone chemical shift data.