ABSTRACT
Building a structural model for every protein from the proteome is an important aspect of understanding protein function and its differences between homologs, variants, mutants, etc. The task of generating hundreds of thousands of models does not seem impossible, due to progress in both experimental structure determination (1) and theoretical structure prediction by homology techniques (2). However, all models necessarily contain an elaborate network of errors of varying character and magnitude, depending on the derivation method. Understanding these errors is key to the utility of a model.
