Models of protein–ligand crystal structures: trust, but verify

Abstract

X-ray crystallography provides the most accurate models of protein–ligand structures. These models serve as the foundation of many computational methods including structure prediction, molecular modelling, and structure-based drug design. The success of these computational methods ultimately depends on the quality of the underlying protein–ligand models. X-ray crystallography offers the unparalleled advantage of a clear mathematical formalism relating the experimental data to the protein–ligand model. In the case of X-ray crystallography, the primary experimental evidence is the electron density of the molecules forming the crystal. The first step in the generation of an accurate and precise crystallographic model is the interpretation of the electron density of the crystal, typically carried out by construction of an atomic model. The atomic model must then be validated for fit to the experimental electron density and also for agreement with prior expectations of stereochemistry. Stringent validation of protein–ligand models has become possible as a result of the mandatory deposition of primary diffraction data, and many computational tools are now available to aid in the validation process. Validation of protein–ligand complexes has revealed some instances of overenthusiastic interpretation of ligand density. Fundamental concepts and metrics of protein–ligand quality validation are discussed and we highlight software tools to assist in this process. It is essential that end users select high quality protein–ligand models for their computational and biological studies, and we provide an overview of how this can be achieved.

Molecular dynamics simulations reveal structural insights into inhibitor binding modes and functionality in human Group IIA phospholipase A2

“…Focused scrutiny of the ligands, and the electron density in the vicinity, can remove most ambiguity, though assessments of the interpretations are sometimes continuously made, including after deposition of coordinates has been made with the databank . The trials and tribulations of placing ligands in protein crystal structures have been reviewed in the literature from a number of perspectives . There is also evidence that the details of the crystal, and the manner in which the ligand is introduced can affect the mode of binding, representing strong evidence that there is no complete guarantee that publically released structural coordinates are representative for the biologically active species.…”

“…48 The trials and tribulations of placing ligands in protein crystal structures have been reviewed in the literature from a number of perspectives. 47,[49][50][51][52] There is also evidence that the details of the crystal, and the manner in which the ligand is introduced can affect the mode of binding, representing strong evidence that there is no complete guarantee that publically released structural coordinates are representative for the biologically active species. Very recently a report has been made that the existence of polyethylenes as precipitant or cryoprotectants in crystals can influence the location observed of ligands, 53 and although there are no polyethylenes in any crystal structures used in this study, it serves as a reminder that variants of the protein environment could have an effect on the location of the ligand.…”

Modeling zinc‐oxygen coordination in histone deacetylase: A comparison of semiempirical methods performance

“…All the crystallographic structures selected for this work have X‐ray resolutions, being lower than 2Å, with the exception of 5EFB, which has a resolution value of 2.54Å (the resolution values of each structure are presented in Tables and ). These values, together with the validation report data, such as the real space correlation coefficient (RSCC), in the case of the ligands present in the structure, are related to the spatial position accuracy of the atoms . The RSCC indicates how faithful a ligand is inserted to its electronic density, in which values close to 1 indicate a perfect fit into the electronic density .…”

Knowledge-Based Strategy to Improve Ligand Pose Prediction Accuracy for Lead Optimization

“…In order to produce statistically significant comparisons, we only considered proteins with more than 20 ligand pairs with no other filters applied. This automated curation process could include a small number of poorly refined structures, as previously noted by PDB data quality survey 27,28 .…”

“…b) Coverage on active compounds (< 500 nM in main in vitro assay) only. 28 In this article we presented a validation and analysis of a maximum common substructure- 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 89x66mm (300 x 300 DPI) …”