The data collected on the X31 beam line was phased using the CCP4 program MLPHARE  and the initial phases subsequently subjected to electron density modification techniques implemented in the program SQUASH. A standard procedure was adopted for the phasing which is summarized as follows
Several electron density maps were produced with the above procedure using several combinations of data sets. Maps were also calculated using all six of the known Iridium sites for phasing and also only the four sites given by MULTAN to see whether this solution would have sufficed for structure determination. Map correlation coefficients and average phase deviations for the six maps were calculated against the refined derivative structure as described in Sec. . A total of six maps were calculated from the MAD data using the following approaches -
Phasing was also performed using the SIRAS method. The native data plus data sets 3, 4 and 5 from the derivative crystal were used. Map correlation coefficients and average phase deviations for the SIRAS phasing results were calculated against the refined native structure. The results from the each of the above calculations are summarised in Table . The program MLPHARE allowed the option of assigning anisotropic temperature factors to the heavy atoms and to refine the temperature factors as such. However when this was attempted no improvement in the electron density maps was observed compared with the use of isotropic temperature factors.
Table: Result of phasing with six combinations of data from different energies. Mean phase deviations and correlation coefficient are shown for the maps produced by phases from MLPHARE and the phases after running SQUASH. Also shown are the mean figures of merit for the MLPHARE phases.
The figures of merit for the centric and acentric reflections produced by MLPHARE appear to be consistent with the idea that the addition of extra information into the phasing procedure increases the quality of the resulting phases. The lowest FOM's are observed for phasing method 6 where only two MAD data sets are used and the highest are observed for the SIRAS phasing. The map correlation coefficients and phase deviations are however not fully consistent with these results. The SIRAS phasing results are in fact the least correlated with the `true' target structure and have also the largest phase deviations. Further more solution 6 gives the lowest mean phase deviations.
Figure shows a break down of the phase deviations as a function of for each solution. From this graph it is apparent that solution 6 has lower phase deviations for lower resolution data where Å. Solutions 3, 4 and 5 are of similar quality and follow similar trends as a function of . These three solutions give the highest map correlation coefficient, each greater than .
Figure: Breakdown as a function of of the average phase deviation between different MLPHARE phasing solutions and the refined equivalent structure solution.
The application of the program SQUASH resulted in a significant improvement in the map correlation coefficients of all the electron density maps obtained by the MAD method. This was not the case for the SIRAS solution however.
Examples of the electron density maps produced by the MAD phasing solutions 3 and 5 and by the SIRAS method are shown on the following pages together with the electron density maps obtained using phases and structure factors obtained from the refined derivative structure for comparison. The maps are contoured at Rms density. Figs. shows one of the three helices in the lysozyme structure which extends from Thr 89 to Lys 96. Figs. shows a region of strand running from Thr 40 to Tyr 53. Figs. shows the single Trp 28 residue. These three map regions were chosen since they represent the most typically recognisable features of an electron density map. When setting out to build a protein model into an experimental electron density map they act as good starting points for map interpretation. An important feature of an experimental density map is its connectivity, i.e. the continuity of the electron density along the main chain. Good connectivity allows large regions of the protein main chain to be built by partially automated methods.
Inspection of the electron density maps confirms the indications given by the map correlation coefficients. It is believed that even the electron density map produced by solution 6 using only two data sets may have proved to be interpretable. There is however a clear improvement to be achieved by including a third data set as in solutions 1, 3 and 4. The addition of the two extra iridium sites into the phasing resulted in a small improvement in the maps but the solutions using only the four sites produced by MULTAN still appear to give maps of acceptable quality.
Figure: Alpha helical region extending from residue 87 to 101. Electron density maps for the refined derivative structure and MAD phasing solutions 3 (using data sets 3, 4 and 5) and 5 (using all five data sets measured on X31) are shown. Also shown is the result of phasing using the SIRAS method. The stick model is coloured as follows: yellow:carbon; blue:nitrogen; red:oxygen; green:sulphur. The electron density map is contoured at Rms density. (The diagram was produced using O, ODLEDIT and OPLOT as were all the following electron density maps.)
Figure: Beta strand region extending from residue 40 to 53. Electron density maps for the refined derivative structure, MAD phasing solutions 3 and 5 and the SIRAS solutions are shown.
Figure: Trp 28 residue. Electron density maps for the refined derivative structure, MAD phasing solutions 3 and 5 and the SIRAS solutions are shown.