Assessing the applicability of Bayesian inference for merging small molecule microED data

Microcrystal electron diffraction (MicroED) is an emerging technique for characterizing small molecule structures from nanoscale crystals. Merging data from multiple crystals is a particularly challenging step in the microED workflow. A common practice is to manually curate datasets and apply scaling programs conventionally utilized in rotational X-ray diffraction (XRD), but this could be time-consuming and risks introducing human bias in data analysis. Recently, a Bayesian inference program named Careless (Dalton et al., 2022) has demonstrated excellent performance in merging macromolecular XRD data. Here, the applicability of Careless to small molecule microED data is evaluated and an investigation of the impact of dataset curation is performed. Benchmarking against XDS/XSCALE shows that Careless is an effective complementary approach that merges data to a higher CC1/2 value at high resolution. Furthermore, merging outcomes are not significantly improved by curating datasets either manually or with an automated extension to Careless, cautioning against the common practice of manual dataset curation.