LS-CAT

SynchWeb / ISPyB Guide

Automated Processing & Reprocessing

ISPyB automatically kicks off data reduction as your images come in, and gives you tools to dig into — or redo — that processing.

Autoprocessing pipelines

Several pipelines can run automatically on a completed data collection:

• Fast_DP
• Xia2 (3dii)
• Xia2 (DIALS)
• Xia2 (Multiplex)
• autoPROC / autoPROC & STARANISO

Confirm with LS-CAT staff which of these pipelines are currently enabled on your local ISPyB instance — the set deployed at LS-CAT may differ from this list.

A gear icon shows while a pipeline is running. Once finished, the results bar expands into a panel with a tab per pipeline, showing resolution range, spacegroup, signal-to-noise (<I/σ(I)>), R_meas, completeness, cell parameters, beam centre, multiplicity, anomalous completeness, and CC½.

Phenix xtriage

phenix.xtriage runs automatically on autoprocessed results and flags potential problems with green (ok), yellow (alert), and red (serious warning) messages — useful for spotting issues that might be fixed by adjusting your collection strategy or reprocessing with different settings. Find the xtriage log through Logs & Files for that dataset. See the Phenix xtriage documentation for how to interpret specific warnings.

Logs and files

Each autoprocessing pipeline has a Logs & Files button with everything it produced — HTML reports (often with embedded graphs), JSON data, and plain text logs. Files open in a viewer inside the page; the xia2 report is a good example.

If you need something not shown there, click the triangle icon on a dataset to reveal the raw processing path on disk, generally following a pattern like /<lscat-data-root>/<beamline>/<year>/<proposal>/processed/<protein>/<sample>/<dataset>/<pipeline>/ — confirm the exact convention with LS-CAT staff. Processed files are generally kept on disk longer than raw images.

An archive button downloads a zipped copy of a pipeline's output. Result-type files can also be pushed directly to CCP4 Cloud, if you've configured your CCP4 Cloud account and Cloudrun ID in SynchWeb.

Data quality plots

From an autoprocessing pipeline's overview, the Plots button compares resolution against metrics like CC½ or number of observations. You can plot multiple runs against each other — different crystals, different sweeps, different pipelines (e.g. DIALS vs. xia2) — and add a second parameter on its own scale for direct comparison.

Radiation damage plot

Generated alongside Fast_DP results. Open a Fast_DP tab and choose radiation damage from the side menu. The plot follows the method in Diederichs, Acta Cryst. (2006) D62, 96–101: a roughly flat R_d-vs-frame line means minimal damage; a rising slope points to progressive radiation damage, often mirrored by a falling spot count in the per-image analysis plot.

Reciprocal space viewer

From a dataset's Attachments, open the file view to launch a UglyMol-based reciprocal space viewer — a quick way to sanity-check diffraction spot quality without leaving the browser.

Downstream processing

Downstream processing picks up after autoprocessing hands off a scaled, merged MTZ file, and covers phasing and initial model building. It needs a bit more from you first: a protein sequence, anomalous scatterer information (if relevant), and/or a PDB file on the protein record.

Results appear as expandable tabs per dataset with refinement graphs, heavy-atom site tables, and an interactive map/model viewer (UglyMol). BigEP is a convenience wrapper that kicks off autoSHARP, phenix.autobuild, and Crank2 together.

AlphaFold-assisted processing

Confirm with LS-CAT staff whether AlphaFold-assisted model generation is enabled on your local ISPyB instance before pointing users to this section.

Where enabled: upload a single-chain protein sequence to ISPyB, ideally before you ship, to trigger AlphaFold model generation. The resulting ranked PDB models feed into DIMPLE and MrBUMP after data collection alongside any homology or user-supplied models. You can also upload your own predicted or experimental models (AlphaFold/EBI, RoseTTAFold, existing PDB entries) for use in molecular replacement. If you use AlphaFold results, remember to cite Jumper et al. (2021, Nature), the source database, and the autoprocessing/downstream tools you used.

Reprocessing data

SynchWeb lets you resubmit a dataset to LS-CAT's processing cluster with different settings. Reprocessing works on data still available on disk — see LS-CAT's data retention policy for how long that is guaranteed. Older data may need to be retrieved first; contact LS-CAT staff for help.

1. Click the gear icon on a dataset to open reprocessing options.
2. Choose a pipeline (XDS 3dii or DIALS via xia2, Fast_DP, autoPROC).
3. Set a high-resolution cutoff, spacegroup, and cell parameters if needed.
4. Pick an image range — by start/end fields or interactively on the spot-analysis view — useful for cutting out radiation-damaged frames.
5. Click Integrate to submit the job to the cluster.

Results show up as an additional tab alongside the original autoprocessing. For multiple sweeps or crystals of the same protein, the multi-xia2 pipeline can process them together; results are available both through the reprocessing tab and on disk.