Use of noncrystallographic symmetry for automated model building at medium to low resolution

Abstract

A novel method is presented for the automatic detection of noncrystallographic symmetry (NCS) in macromolecular crystal structure determination which does not require the derivation of molecular masks or the segmentation of density. It was found that throughout structure determination the NCS-related parts may be differently pronounced in the electron density. This often results in the modelling of molecular fragments of variable length and accuracy, especially during automated model-building procedures. These fragments were used to identify NCS relations in order to aid automated model building and refinement. In a number of test cases higher completeness and greater accuracy of the obtained structures were achieved, specifically at a crystallographic resolution of 2.3 Å or poorer. In the best case, the method allowed the building of up to 15% more residues automatically and a tripling of the average length of the built fragments.

Figure 1

More than 50% of all structures in the Protein Data Bank (PDB) have NCS relations (derived from PDB data, March 2011).

Figure 2

The result of standard automated protein model building of the test case shiga-like toxin (PDB entry 1c48) with ARP/wARP and X-ray data to 3.0 Å resolution. No NCS was used and each of the five subunits was built in a different way.

Figure 3

Workflow of the Protein NCS-based Structure Extender (PNS Extender). Intermediate partial models are examined for symmetric dependencies between stretches of two fragments (a). An initial match is found between green and blue regions (b, red blocks). The initial match is extended in both directions of the chain fragments (c, orange blocks). Once the extension is finished and the r.m.s.d. between the extended matches (red blocks in d) is still below the acceptance threshold, each extension (e, overlayed blocks) is NCS-transformed, as shown by arrows, onto the other fragment. Finally, longer extended green and blue fragments are obtained (f) and their extended parts (f, yellow blocks) are input as guides for protein-chain tracing.

Figure 4

Flowchart of ARP/wARP protein model building, including automatic NCS detection and its use for extension of the model.

Figure 5

Estimated reliability of the derived weights used for NCS extension and the accuracy of the obtained extended parts of the model.

Figure 6

Validation of method properties: exclusion test. (a) The original structure (PDB entry 1c48); (b) the same structure with 35% of all residues excluded. In (c) all missing residues in (b) were retrieved using PNS Extender.

Figure 7

The best results for tests with variable r.m.s.d. thresholds for acceptance of identified NCS matches (0.4/0.5 Å) and a variable amount of top-ranking fragments to be fed back into the model-building process. (a) Average completeness of the built model. (b) Average length of built fragments. (c) Residues that have been assigned to sequence.