journal menu
![[Figure 6]](qh5069fig6mag.jpg)
|
|
Figure 6
The dependence of merging success on phase accuracy, completeness and the angular spread of reflection data. A total of 25 920 simulated data sets were generated across three crystal systems, spanning a range of P1 completeness, phase errors and relative B factors. Pairs of data sets with similar starting characteristics and in random orientations were merged. (a) The fractional merge error was computed as the magnitude of the vector error between the true and estimated fractional shifts required to position data sets on a common phase origin. This metric is shown as a function of the mean phase error (left), P1 completeness (center) and how unevenly spread the reflections were across the tilt range (right) for the data sets being merged. The latter was estimated as the Jensen–Shannon (JS) distance to a uniform angular distribution spanning ±60° (see equation 8 ![[link]](../../../../../../logos/arrows/d_arr.gif) ). Although 2880 pairs of data sets with a phase error of 0° were merged, the results are visually superimposed in the leftmost panel. Merge errors are shown as a function of (b) these three parameters or (c) the two indicated parameters of the data sets being merged. In (b) and (c), red and black indicate data sets for which the incorrect and correct origin shift, respectively, was determined by the merging algorithm. The error threshold was set to be twice the sampling interval used by the algorithm, such that the correct phase origin was found only when the fractional merge error was <0.03. |
![[Figure 6]](qh5069fig6.jpg)
 | STRUCTURAL BIOLOGY |
ISSN: 2059-7983
access
