Research Article: Substructure determination using phase-retrieval techniques

Date Published: February 01, 2018

Publisher: International Union of Crystallography

Author(s): Pavol Skubák.


The relaxed averaged alternating reflections (RAAR) phase-retrieval method has been applied to crystallography for the first time and has been shown to outperform charge flipping in anomalous substructure determination.

Partial Text

Rapid progress in both instrumentation and computational methods of macromolecular imaging has led to unprecedented growth in the number of macromolecular structures solved: the number of structures deposited in the Protein Data Bank (PDB; Berman et al., 2000 ▸) has increased by an order of magnitude in the new millennium, with the majority of these PDB entries being solved by X-ray crystallography. Owing to the rapidly growing number of known structures, molecular replacement (MR), a technique to determine the structure under study using similar previously determined folds, has become the most frequently used technique to solve the phase problem in macromolecular X-ray crystallography: over two thirds of the X-ray crystallographic structures deposited in the PDB were solved by MR or by a combination of MR with experimental phasing techniques.

Fig. 2 ▸ shows the performance of PRASA in terms of substructures determined and macromolecular models built for the 169 SAD data sets. Owing to the ability of the ‘combined’ building algorithm to complete partial models, almost all of the resulting models can be divided into two distinct categories: either correctly built close to completion (more than 75% of the backbone correctly traced) or not built (less than 25% of the backbone correctly traced). As can be seen from Fig. 2 ▸(b), three models fall outside these categories: in two cases the limiting factor behind the partial (50 and 69% complete) models was the low resolution of the data set (3.88 and 3.2 Å, respectively), while the remaining data set, which was built to 59%, suffered from twinning. For the sake of simplicity, the few partially built models will be considered as correctly built in the following text.

In the tests on 169 SAD data sets, it has been shown that the RAAR algorithm, implemented in the new program PRASA for substructure determination, outperforms the charge-flipping algorithm as implemented in the same program. An analysis of the anomalous signals of the data sets solved only by RAAR indicates that the RAAR algorithm extends the limits of charge flipping towards data sets with weaker anomalous signals.




Leave a Reply

Your email address will not be published.