cctbx.xfel GUI/Run Stats tab

The Run Stats tab is the live view of how processing is going: for each image of the selected runs it shows how many strong spots were found, whether the image indexed, to what resolution, and the photon energy, together with per-run hit and indexing rates. It is the tab to keep open during data collection. The plot is redrawn every 5 seconds by the Run Stats Sentinel while Auto update is ticked.

Back to cctbx.xfel GUI.

Selecting what to plot

The Statistics Options box on the lower left:

• Trial: which trial's results to show. The list includes every trial; pick None to clear the plot.
• Selected runs: tick the runs to plot. The list contains every run in the trial's run blocks. Ticking runs by hand turns off the two automatic selections below.
• Auto plot last five runs: keep showing the five most recent runs in the trial, following along as new runs arrive. This is the normal setting during collection.
• Auto plot entire experiment: show every run in the trial in a compact three-panel form (no per-run annotations or energy panel), for an overview of a long experiment.
• Hide options collapses the options and image lists to give the plot the whole window.
• Auto update: untick to freeze the plot (and stop the sentinel), for example to zoom in with the matplotlib toolbar without the view being reset.

The threshold fields take effect when you press Enter in the field:

Field	Default	Used for
High resolution limit	2.0 Å	An indexed image counts as high quality if its high-resolution bin (where I/σ(I) drops below the I/σ(I) cutoff) reaches this resolution. The orange line in panel 3 and the parenthesised counts in panel 4 use it.
# multiples threshold	2	An image with at least this many lattices counts as a multi-lattice hit (pink line).
two theta ratio cutoff	1	An image whose radial-average ratio at the run block's two 2θ angles exceeds this counts as a solvent hit (green line), that is, a shot that hit the liquid stream or drop whether or not it contained a crystal.
# strong spots cutoff	16	An image with at least this many strong spots counts as a crystal hit. Also defines which images go into the two image lists below.
I/sig(I) cutoff	1	Defines the high-resolution bin of each indexed image.
# images to dump	10	How many images to convert and open from each list (LCLS XTC processing only).

Reading the plot

The x axis is the image timestamp, with runs laid end to end and separated by vertical lines. Each run is labelled at the top with its tags and its job status, coloured blue for DONE, green for running or queued, red for failed and black if no job was found.

1. Strong spots: one point per image, the number of strong spots found. Blue points indexed, grey points did not. The y axis is clipped at 5000.
2. Rates (running averages over a window of images): blue is the percentage of images that indexed, pink the percentage with multiple lattices, and green (right axis) the percentage of solvent hits.
3. Resolution: orange points give each indexed image's high-resolution bin in Å; the orange line (right axis) is the percentage of images whose resolution reaches the high resolution limit.
4. Energy: the photon energy of each image in eV, with a text summary per run: run number; images collected / crystal hits; images indexed (indexed to the resolution limit); percentage solvent hits / crystal hits; percentage indexed (to the resolution limit).

The Large text toolbar button enlarges the markers and labels.

Clicking on the plot

At LCLS, clicking on a point in the strong spots or resolution panel opens that image in the DIALS image viewer, with its strong spots and, if it indexed, its indexed reflections overlaid. The viewer stays open and loads a new image on each click. The matplotlib toolbar's zoom and pan modes must be off for clicks to be interpreted this way.

Image lists (LCLS XTC processing)

When the processing back end is cctbx.xfel.xtc_process, two lists appear on the right:

• Strongest Indexed Images: images that indexed and have at least # strong spots cutoff spots, with their file paths. Open images opens the first # images to dump of them in dials.image_viewer with their indexed reflections (converting them from the XTC stream first if they were not dumped).
• Strong Images that Didn't Index: images with many strong spots that failed to index, the ones to look at when the indexing rate is lower than the hit rate suggests it should be. Dump images extracts them from the XTC stream with cctbx.xfel.xtc_dump into the job's all/ folder and opens them in the image viewer.

Both commands run in the background; their output appears in the terminal. With the standard cctbx.xfel.process back end these lists are hidden; use the Dump all images to disk option in LCLS options and look in the all/ folder instead.

Typical use

• During collection: select the current trial, press Auto plot last five runs, and watch the hit rate (grey and blue points, green line) to judge sample delivery, and the indexing rate (blue line) and resolution (orange) to judge data quality.
• Tuning spot finding: create a trial with Index off, and compare the strong-spot counts of a few runs between trials with different spot finding thresholds. Then open the strongest images to confirm that the spots found are real.
• Diagnosing indexing: if many images have plenty of strong spots but did not index (grey points high in panel 1), check the unit cell and space group in the trial, the detector distance and beam centre in the run block, and look at a few of those images.