Schematic workflow of data assembly. Firstly, individual microcrystal data sets (100 frames each) were split and processed into ten accumulated wedges; and wedges with maximum CC1/2 were selected for data assembly. Secondly, iterative crystal rejection was performed by rejection of 100 microcrystals with the highest average SmRmerge, 〈SmRmerge〉. After each crystal rejection, frame rejection was performed to remove outliers based on defined SmRmerge cutoffs. Finally, the assembled data set with maximum ACC was used for structure determination.