Fig. 4

A simple protein preparation pipeline. In working with protein structures, e.g., to create the Prop3D-20sf dataset, each domain is typically corrected or ‘sanitized’ by adding missing atoms and residues, checking rotameric states (highly-populated rotamers should be assigned, by default), protonating, and performing a crude potential energy minimization of the 3D structure; this general workflow is sketched here using a tripeptide segment (PDB entry 1KQ2)