Lysozyme revisited: evaluating models of a reference protein in structural biology

We have evaluated the quality of over 1200 crystal structures of hen egg-white lysozyme (HEWL) deposited in the Protein Data Bank (PDB). These structures, collected over nearly 50 years, vary in quality, despite all representing essentially the same small enzyme consisting of 129 amino acid residues. Some of the entries originated from studies of the binding of small-molecule ligands to HEWL, whereas the majority of deposits represent the outcomes of tests of new experimental approaches to crystallization and data collection and/or evaluations of new computational protocols. We found no correlation between R _free, which is a measure of structure quality, and R _merge, an indicator of raw data quality, for 136 near-atomic-resolution lysozyme structures. We found out that many of the lysozyme structures deposited as a result of methodology evaluation are not fully or correctly refined. We, therefore, propose that such structures be appropriately flagged in the PDB with a CAVEAT record to prevent their inadvertent inclusion in large-scale data mining analyses or training sets for artificial intelligence methods.