Breaking the indexing ambiguity in serial crystallography

In serial crystallography, a very incomplete partial data set is obtained from each diffraction experiment (a 'snapshot'). In some space groups, an indexing ambiguity exists which requires that the indexing mode of each snapshot needs to be established with respect to a reference data set. In the absence of such re-indexing information, crystallographers have thus far resorted to a straight merging of all snapshots, yielding a perfectly twinned data set of higher symmetry which is poorly suited for structure solution and refinement. Here, two algorithms have been designed for assembling complete data sets by clustering those snapshots that are indexed in the same way, and they have been tested using 15 445 snapshots from photosystem I [Chapman et al. (2011), Nature (London), 470, 73-77] and with noisy model data. The results of the clustering are unambiguous and enabled the construction of complete data sets in the correct space group P6(3) instead of (twinned) P6(3)22 that researchers have been forced to use previously in such cases of indexing ambiguity. The algorithms thus extend the applicability and reach of serial crystallography.