An Iterative Direct Phasing Algorithm for Protein Structure Determination

An iterative phasing method based on the hybrid input-output (HIO) algorithm, proposed by He & Su (2015), is improved to retrieve phases of protein crystals. By testing the method on a protein crystal with 65% solvent content, it is proved that this method is capable of phasing protein crystals with an intermediate solvent content. Results from a blind test have yielded an interpretable density map, which shows that the phasing method is capable of solving completely unknown protein structures. The phasing method was tested on medium- and low-resolution diffraction data of a membrane protein, which shows that the atomic structure can be solved directly from the diffraction data with resolution higher than 3.5Å. For data of lower resolutions, secondary structures and the protein boundary can still be obtained. The results also indicate that there is an optimal histogram to use for data at various resolutions. By averaging over tens of successful runs, the resulting density map can be greatly improved and the final phase error can be reduced by about 10 degrees, thus leading to a better reconstructed model. Finally, a density constraint called non-crystallographic symmetry (NCS) is employed to overcome the undersampling condition of protein crystals with low solvent content. The phasing method has been applied to a protein crystal with 3-fold rotational NCS and 45% solvent content. The results show that, when the NCS axis is well positioned, the protein envelope and the NCS mask can be automatically rebuilt. By employing NCS averaging, this method is thus capable of solving protein crystals with low solvent content.