“…Even where major conformational differences are absent, circumventing non-isomorphism problems lowers errors in structure-factor measurements, and clustering data sets could provide a valuable means for identifying systematic errors (Diederichs, 2017). The general importance of redundancy in data collection is increasingly being recognized, and serial crystallography experiments are becoming routine Weinert et al, 2017;Mathews et al, 2017;Standfuss & Spence, 2017), but averaging measurements of potentially nonequivalent physical entities (for example structure factors from different regions of a crystal) may be detrimental. The improvement in data quality that can be gained by clustering multiple related data sets into isomorphous groups during data reduction has been explored previously in several contexts (Liu et al, 2011;Giordano et al, 2012;Foadi et al, 2013;Zander et al, 2016;Assmann et al, 2016;Diederichs, 2017;Yamamoto et al, 2017), although previous work has generally focused on non-isomorphism resulting from changes to crystal-packing interactions, rather than emphasizing the potential to use this feature of protein crystals as a means to explore conformational heterogeneity.…”