“…Decoupling these two types of errors and discarding datasets with large systematic errors (non-isomorphous) is not trivial. Various concepts have been developed for crystallographic data selection, scaling and merging such as hierarchical cluster analysis (HCA) based on cross-correlation among datasets (Giordano et al, 2012;Santoni et al, 2017), unit-cell-based hierarchical clustering (BLEND; Foadi et al, 2013), genetic algorithm (CODGAS; Zander et al, 2016) and xscale_ isocluster (Diederichs, 2017). All these concepts have been successfully demonstrated on various cases to provide complete and high-quality datasets by careful data selection and merging.…”