PDB_REDO: automated re-refinement of X-ray structure models in the PDB

Abstract

Structural biology, homology modelling and rational drug design require accurate three-dimensional macromolecular coordinates. However, the coordinates in the Protein Data Bank (PDB) have not all been obtained using the latest experimental and computational methods. In this study a method is presented for automated re-refinement of existing structure models in the PDB. A large-scale benchmark with 16 807 PDB entries showed that they can be improved in terms of fit to the deposited experimental X-ray data as well as in terms of geometric quality. The re-refinement protocol uses TLS models to describe concerted atom movement. The resulting structure models are made available through the PDB_REDO databank (http://www.cmbi.ru.nl/pdb_redo/). Grid computing techniques were used to overcome the computational requirements of this endeavour.

Figure 1

R-free values extracted from the PDB header (diamonds) and values recalculated with Refmac (Winn et al., 2001 ▶) using the deposited experimental data (squares) plotted against the experimental data resolution. The values are averages for all structure models in 0.1 Å bins. The recalculated R values (not shown) follow the same pattern.

Figure 2

R-free values extracted from the PDB header (diamonds) and values obtained after re-refinement in Refmac (Winn et al., 2001 ▶) with TLS models (squares) plotted against the experimental data resolution. The values are averages for all structure models in 0.1 Å bins. The effect of the TLS parameterization is clearly shown by the results of the re-refinement without TLS models (dotted line). For all but the highest resolution bins, refinement with TLS gives lower average R-free values.

Figure 3

WHAT_CHECK Ramachandran plot appearance Z-scores (Hooft et al., 1997 ▶) for original (diamonds) and TLS-refined structure models (squares) as a function of resolution. The values are averages for all structure models in 0.1 Å bins.

Figure 4

Atomic overlaps (bumps) per structure model as detected by WHAT_CHECK for original (diamonds) and TLS-refined structure models (squares) as a function of resolution. The values are averages for all structure models in 0.1 Å bins.

Figure 5

Bond-length r.m.s. Z-score per structure model as calculated by WHAT_CHECK for original (diamonds) and TLS-refined structure models (squares) as a function of resolution. The values are averages for all structure models in 0.1 Å bins.

Figure 6

Bond-angle r.m.s. Z-score per structure model as calculated by WHAT_CHECK for original (diamonds) and TLS-refined structure models (squares) as a function of resolution. The values are averages for all structure models in 0.1 Å bins.

Figure 7

Percentage of structure models that improve in terms of R-free after TLS refinement plotted as a function of the year of deposition. The percentage of all evaluated structures (diamonds) decreases from 90% for 1995 to 62% for 2006. The percentage of structures previously refined with TLS (squares) varies between 21 and 32%.