view article
Figure 8
Indexing quality profiling of TORO (black dashed line) versus XGandalf (red solid line) across the lysozyme and CXIDB ID 83 data sets, exploring varying TORO hyperparameters for optimal speed. Metrics presented include CC versus Resolution, I/σ versus Resolution and Redundancy versus Resolution. Distinct columns represent data from the lysozyme and CXIDB ID 83 data sets, respectively. For context, findings from Fig. 4 for both TORO and XGandalf are juxtaposed against TORO RT (dashed–dotted grey line) – a swifter yet slightly less precise variant of TORO, configured with lattice_size=10000, angle_resolution=100 and num_top_solutions=25. TORO RT boasts a processing rate of 3006.76 images s−1 on an A100 GPU in standalone mode, making it a suitable candidate for real-time feedback indexing. Quality-wise, TORO RT remains robust, matching (lysozyme) or outperforming (CXIDB ID 83) XGandalf. Data sets were processed using indexamajig with --no-retry --no-refine --no-check-cell.

Journal logoJOURNAL OF
APPLIED
CRYSTALLOGRAPHY
ISSN: 1600-5767
Volume 57| Part 4| August 2024|
https://doi.org/10.1107/S1600576724003182
Follow J. Appl. Cryst.
Sign up for e-alerts
E-alerts
Follow J. Appl. Cryst. on Twitter
Twitter
Follow us on facebook
Facebook
Sign up for RSS feeds
RSS