CaBLAM scores were used to guide the enforcement of Ramachandran restraints. At several points during the iterative refinement of the final FL-20Sfocus models, Ramachandran restraints were imposed using the Oldfield target function (Oldfield, 2001) as implemented in phenix.real_space_refine (Adams et al., 2010; Headd et al., 2012). A grid of refinements with different values of the parameters weight_scale and plot_cutoff were performed; the results are visualized as surfaces for three key diagnostic parameters—CCmask, Ramachandran fraction favored, and the total fraction of residues flagged by CaBLAM analysis (i.e., the total fraction of residues flagged as either outliers, disfavored, or severe) (Richardson et al., 2018). The value of plot_cutoff largely sorts results into three matching regions for each metric, while the value of weight_scale is less predictive. The ten refinements with the lowest CaBLAM scores were further examined, and one model was chosen for subsequent manual and automated refinement. The boxes (dashed lines) on the surface plots illustrate the regions from which these models were found in parameter space. Scatter plots reveal the relationships between CCmask, Ramachandran fraction favored, and CaBLAM fraction flagged; the top ten refinements as identified by CaBLAM are shown (red points).