Fig. 5: Careless can leverage geometric metadata to improve XFEL data processing.
From: A unifying Bayesian framework for merging X-ray diffraction data
a By adding image-specific layers to the neural network, Careless can scale diffraction data from serial XFEL experiments. b The Ewald offset (EO), shown in red, can optionally be included during processing. S0 and S1 represent the directions of the unscattered and diffracted X-rays, respectively. c Half-dataset correlation coefficients by resolution bin for data processed in Careless, with and without Ewald offsets, and CCTBX. d Refinement R-factors from phenix.refine using Careless, with and without Ewald offsets, and CCTBX. e Careless 2Fo − Fc electron density map, from inclusion of EO, contoured at 2.0 σ (purple mesh) overlayed with thermolysin anomalous omit map contoured at 5.0 σ (orange mesh). f Peak heights of anomalous scatterers in an anomalous omit map, in σ units, for Careless output with and without Ewald offsets and conventional data processing with CCTBX.
