Detection of unrealistic molecular environments in protein structures based on expected electron densities

Abstract

Understanding the relationship between protein structure and biological function is a central theme in structural biology. Advances are severely hampered by errors in experimentally determined protein structures. Detection and correction of such errors is therefore of utmost importance. Electron densities in molecular structures obey certain rules which depend on the molecular environment. Here we present and discuss a new approach that relates electron densities computed from a structural model to densities expected from prior observations on identical or closely related molecular environments. Strong deviations of computed from expected densities reveal unrealistic molecular structures. Most importantly, structure analysis and error detection are independent of experimental data and hence may be applied to any structural model. The comparison to state-of-the-art methods reveals that our approach is able to identify errors that formerly remained undetected. The new technique, called RefDens, is accessible as a public web service at http://refdens.services.came.sbg.ac.at.

Fig. 1

The environment of Asn-348 in chain C of PDB entry 3fxg, a rhamnonate dehydratase determined at a resolution of 1.9 Å. Atoms are colored by element: carbon, gray; oxygen, red; nitrogen, blue; sulfur, yellow. Contours of constant electron density are rendered as transparent yellow (0.1e⁻) and red (0.2e⁻) surfaces on both sides. The left hand side shows the electron excess as calculated by RefDens. For comparison, the right hand side displays the electron excess according to the measured (Fo-Fc) map from the EDS server. Although derived in substantially different ways, both maps identify the same problem regions

Fig. 2

Impossible crystal contacts for Leu-332 in chain A of PDB code 3f1p, a transcription factor resolved to 1.17 Å. Residue Leu-332 from the asymmetric unit is rendered in dark gray. From a crystallographically related chain another Leu-332, shown in white, occupies the same space as the analyzed residue

Fig. 3

Unrealistically close contact for Lys-A-332 in chain A of PDB code 3gr2, an beta-lactamase resolved to 1.8 Å. The N^ζ of Lys-A-332 is placed way too close to the of Gln-A-361 (1.73 Å). Additionally and of Trp-A-276 are situated at hydrogen bonding distance to the N^ζ atom

Fig. 4

Comparison between RefDens and MolProbity. A single point represents the electron excess on the x-axis and the sum of MolProbity’s clashscores over the atoms of the investigated side-chain. The plot contains scores of 57 erroneous side-chains from our benchmark set for which a clash is detected by MolProbity

Fig. 5

Electron excess as calculated by RefDens for the original PDB entry and the re-refined structure from the PDBRedo webserver. The red lines highlight the 3 e⁻/ų cutoff. A total of 12 side-chain errors are relaxed in the re-refined structures, whereas the other 68 problematic side-chains still exhibit high electron excess

Fig. 6

Interaction of a tyrosine residue with the heme group in chain A of two protein structures of horse heart ferricytochrome c. (Left side) In the solution structure (PDB code 1ocd) RefDens detects a broad region of high electron excess for Tyr-67. Although the axially coordinating methionine Met-80 is close to the tyrosine hydroxyl group, the major source of conflict is the spatially close heme group. (Right side) In the corresponding X-ray structure (PDB code 1hrc at a resolution of 1.9 Å) the heme group is located further from the tyrosine residue and the axial methionine residue binds almost orthogonally to the heme group. This significantly reduces the electron excess such that no conflicting regions are detected for this protein structure

Fig. 7

Unrealistically close contact (2.6 Å) between the S^δ atom of Met-102 in chain A and the backbone N of Ile-103 in PDB entry 2klj, a recently resolved (release date: 2009-10-06) structural protein (left side). In the solution structure RefDens correctly detects a severe electron density overlap. In the corresponding X-ray structure (PDB entry 1h4a, resolved to 1.14 Å) the situation is relaxed (distance of 3.33 Å) which results in a strongly reduced electron density excess (right side)

Fig. 8

Amino acid side-chains rotated and translated into the respective coordinate systems. The regions analyzed for excess electron density are depicted by gray isosurfaces. a Asparagine: The (C^β, C^γ) bond defines the y-axis (green) and both and are placed in the xy-plane (shown as a blue lattice). The C^β atom defines the origin of the coordinate system. The x- and y-axis are shown as red and blue lines, respectively. b Tyrosine: The y- axis is defined by the (C^β, C^γ) bond. The ring system is contained in the xy plane and the C^β atom defines the origin of the coordinate system. c Methionine: The (C^γ, S^δ) bond defines the y-axis and C^ε is placed in the xy-plane. The origin of the coordinate system is given by the center of the C^γ atom