Macromolecular structures: Quality assessment and biological interpretation

Abstract

Structure determination using X-ray crystallography involves collection of diffraction data, determination of initial phases followed by iterative rounds of model building and crystallographic refinement to improve the phases and minimize the differences between calculated and observed structure factors. At each of these stages, a variety of statistical filters exist to ensure appropriate validation. Biologically important observations often come from interpretations of signals that need to be carefully deciphered from noise, and therefore human intervention is as important as the automated methods and filters. Currently, all structural data are deposited in the Protein Data Bank (PDB), and this repository is continuously evolving to incorporate new developments in macromolecular crystallography. The journals that publish data arising from structural studies modulate their policies to take cognizance of new improved methodologies. Together, the PDB and journals have evolved an accepted protocol to ensure the integrity of crystallographic results. As a result, the quality of available data and associated interpretations have improved over the years. Typically, if there are differences regarding the mechanism of action of a protein revealed by crystallography then new experiments are carried out to provide further evidence for or against a particular hypothesis. Hence, the scientific systems in structural biology are robust and to a large extent capable of correcting any errors in interpretation of structural results. However, this process is hindered by spurious challenges to published research, based on misuse of validation mechanisms. Such activities are counterproductive and will cause damage to the field of structural biology. © 2017 IUBMB Life, 69(8):563-571, 2017.

Detect, correct, retract: How to manage incorrect structural models

“…One model [45] was flagged by CheckMyMetal [28] as containing an unusual metal coordination. A study [46] flagged by Salunke and Nair [47] as a case of an electron density map for a ligand contoured at very low level, revealed amongst other errors a different ligand (buffer) in the electron density, which unsupervised methods failed to detect due to an acceptable real space fit and the fact that free amino acid ligands are annotated by the PDB with the same identifier as an amino acid residue and not recognized as a ligand. The example from a recent study [48] was identified based on the correlation emerging between detected errors and certain author clusters.…”

“…Occasionally, however, the critiqued authors respond by pointing out that similar procedures and results were reported before, sometimes by their critics. For example, Table 1 of [47] lists a dozen structures that were accompanied by publications in which F o - F c omit maps showing putative ligand density were contoured at levels lower than 3σ. We analysed the six structures with maps contoured at 1.5σ and concluded that such low level was completely unnecessary for 3r6r [89] and 1ujj [90], since appropriate contour levels show unambiguously the presence of modelled ligands.…”

“…We have pointed out the need for contextual evaluation already in our own analysis of PDB metrics [4, 22]. The contradiction between claim stated and evidence required is particularly concerning for example when the stringent necessity of evidence for an immunogenic peptide in a complex structure [42] is presented on par with modelling of unknown ambiguous solvent density [47].…”

Multiple target sites for designing candidate drugs

“…This is unlike the physiological complexes, which may bind with relatively high affinity with perfect occupancy. It is important to note that if one applies screening criteria based on the high-resolution structures of complexes of high-affinity ligands binding to proteins, such low-affinity leads will certainly be rejected with loss of promising starting compounds [5]. This would be disadvantageous to the very process of drug discovery.…”

Molstack—Interactive visualization tool for presentation, interpretation, and validation of macromolecules and electron density maps

“…In any case, the publication of a re‐refined, corrected model is sometimes a complicated ordeal . In addition, it has also been argued by some authors that the different interpretations stem from the inability of the researchers to reproduce the same electron density maps that were used by the original authors …”