Global vs. Local assessment

When evaluating the quality of a macromolecular model, it is important to consider both the overall structure and specific parts within it. This leads to two complementary types of assessment: global and local.

• Global assessment methods provide an overall measure of a macromolecular structure’s quality. They typically yield a single score or a set of scores that reflect the model’s general agreement with the underlying experimental data (e.g. R-free in X-ray crystallography) and how well it conforms to fundamental stereochemical and physical principles  (e.g. Ramachandran plot). This tells you if there are overall concerns, but not necessarily where the specific problems are.
• Local assessment focuses on specific parts of the model, such as individual amino acids, a ligand binding site, or a flexible loop. Local assessment often involves visually inspecting how the model fits the experimental data in that specific area and checking local geometry. It helps identify areas of high or low confidence that might not be apparent from looking only at global metrics. Local assessment allows you to pinpoint problematic areas that might need closer inspection or indicate functionally flexible regions.

Both global and local assessments are essential for a comprehensive evaluation of macromolecular structures. A model with good global scores might still have local issues in a region critical to your research. Conversely, a structure with some global “red flags” might still have a reliable core relevant to your work. Global assessment is your starting point to get a general sense of confidence before you examine specific regions in more detail.

It is crucial to remember that structural models are interpretations of experimental data. The process of building and refining models involves some degree of subjective decision-making. Also, structure models are made by different researchers using different tools at different times. Methods for sample preparation, data collection, and model building have all improved over the years. Some models are made by experienced researchers with plenty of time, whereas other models are made by young researchers under enormous time pressure.

Furthermore, no single, simple metric exists that can perfectly capture and reflect all parameters of an atomic model, its agreement with diverse experimental data, and its adherence to all chemical environment principles simultaneously. The definition of what constitutes a “bad” structure depends entirely on how you intend to use the model.