Validating protein structures fit into cryoEM density maps
Increasingly, cryo-electron microscopy reconstructions are of suitable resolution to permit tracing protein backbones and identify sidechain conformations. For evaluating atomic models constructed and refined into such maps, it is critical to have some validation metric for assessing the fit of the models to the data. Such a metric would provide a means to determine the proper balance between the fit to the density and model energy and stereochemistry during refinement, and is likely to be useful in determining values of model building and refinement metaparameters quite generally.
Our recent manuscript develops such a measure. Using model-map agreement to an independent reconstruction in high-resolution shells, we show that this independent measure may detect model overfitting to cryoEM density. By adjusting the weight w_a controlling the scaling of physicochemical energy versus experimental energy, we find a clear optimum value; above 0.1–0.2 there is an “inflection point,” where the fit of the refined models to the training map continues to increase but the fit to the testing map decreases, suggesting overfitting to the training map.