Computational protocol: Selenium single-wavelength anomalous diffraction de novo phasing using an X-ray-free electron laser

Similar protocols

Protocol publication

[…] The structure was solved with a combination of the PHENIX and CCP4 suites using a combinatorial grid search of density modification parameters. The SFX structure was solved by first converting the output MTZ reflection file from CrystFEL-0.6.2 (ref. ) into the CCP4 suite using the programs POINTLESS, AIMLESS and TRUNCATE. This was necessary to place the Friedel opposite reflections into the conventional asymmetric unit as well as to adjust the estimated standard deviations; an error scale factor between 0.4 and 0.9 was necessary for subsequent phasing calculations to succeed. ‘Success' was defined as the eventual discovery of at least one non-crystallographic symmetry (NCS) operator in the density-modified map by phenix.find_ncs. The four selenium sites were found using phenix.hyss, and heavy-atom phase probabilities computed by one cycle of the CCP4 program MLPHARE. Any cycles of refinement in MLPHARE were found to degrade the phasing signal. The Hendrickson-Lattman coefficients obtained were then scaled up by a factor of 3–6 using the CCP4 program SFTOOLS. This ‘sharpens' the phases, or reduces the estimate of phase error supplied to the density modification step. This phase sharpening increased the final success rate by 20-fold over the unsharpened phases. Density modification was carried out using the CCP4 program DM, with the majority of successes found after running 300–700 cycles of solvent flattening and histogram matching with the default automatic weighting scheme. After this first round of density modification, phenix.find_ncs was run to search for a (NCS) operator. The success of NCS operator discovery was found to be a powerful indicator of downstream phase quality, and served as an excellent filter for error scaling and phase sharpening parameters. Once discovered, the NCS operator was applied for a second round of density modification, starting over with the same sharpened MLPHARE phases, but this time adding NCS averaging with a spherical mask centered on the ‘guessed molecular center.' Phases of sufficient quality for automatic model building were obtained with spherical averaging mask radius ranging from 5 to 40 Å and auto-updating the mask every 2–50 cycles of DM. Starting with an automatically-derived mask never produced useful phases. In total, over 200,000 DM trials were run, and when these trials were filtered for those where an NCS operator was found and subsequently sorted on the figure of merit from density modification, the trial with the highest figure of merit corresponded to phases of sufficient quality to automatically build the structure. Therefore, although this phasing strategy is not the default in any extant automation package, it does represent a robust phasing strategy that leads to success when the trials are scored by readily available metrics. shows a plot of the correlation of the final maps with the known structure and anomalous peak/root-mean-square (RMS) ratio as a function of the percentage of the total data used.Of the 159,101 jobs run, 178 reported a final figure of merit >0.88, all of which produced high quality maps suitable for automated model building (). The success of NCS operator discovery was found to be a powerful indicator of downstream phase quality, and served as an excellent filter for error scaling and phase sharpening parameters. The final structure produced an Rwork=16.6% and Rfree=19.9% and showed the presence of strong Se peaks, as shown in and as a stereo image in . [...] Crystallographic refinement of the model was performed with the program PHENIX. All the models were checked and completed with COOT. Final model of structure contain residues 16–134 in chain A, B, C and D respectively. Residues (48–50) of chain D were disordered and not included in the final structure. The stereochemical quality of the structure was assessed with PROCHECK. The Ramachandran statistics (most favoured/additionally allowed/generously allowed) are 89.3/10.5/0.2%. The refinement statistics are summarized in . All X-ray crystal-structure figures were produced with PyMOL (http://www.schrodinger.com/pymol). […]

Pipeline specifications

Software tools PHENIX, CCP4, Coot, PROCHECK, PyMOL
Applications Small-angle scattering, Protein structure analysis
Organisms Dipturus trachyderma
Chemicals Selenium