Cheat sheet

XDS processing workflow. The links given like e.g. XDSGUI#Frame refer to the XDSwiki at https://wiki.uni-konstanz.de/xds/index.php/	The programs must have been properly installed - see Installation. Troubleshooting hints are at Problems.
Action	why? what to look out for? what else to know?
1. open a terminal and move the window to the left side of the screen. If you work with HDF5 data, now type export NEGGIA_PATH=/usr/local/lib64/dectris-neggia.so - the path must match YOUR computer; the path given is just a typical one. Type xdsgui and move its window so that you still see the left half of the terminal window.	Keeping the terminal window to the left of the XDSGUI window makes it possible to see some screen output of XDSGUI, e.g. error messages, or the pointless output. For HDF5 data from DIAMOND, use export DURIN_PATH= /usr/local/lib64/durin-plugin.soinstead (of course with the correct path). The actual path must appear in "Generic frame library" under xdsgui / Preferences ../ Paths on macOS, or Menu / Settings / Paths on Linux, and should appear after LIB= in XDS.INP (below). Optionally, before typing xdsgui : cd to an existing XDS directory, or create a new (empty) directory:cd <toplevel-directory>;mkdir <xds-directory>; cd <xds-directory>
2. click the Projects tab and either choose an existing entry (if there is a list of previous projects) or browse to an existing XDS directory, or create a new (empty) directory	If XDS's files already exist in the XDS directory, their content is shown in the tabs of XDSGUI. Note that the XDS directory name appears in the title bar of the XDSGUI window!
3. click the Frame tab and load a frame of your dataset (for HDF5, load xxx_master.h5). Click generate XDS.INP - this reads the header of that frame, and counts the frames of the dataset. Use Zoom, Contrast and Brightness and move around the frame to evaluate the shape and separation of the reflections: are they smeared or sharp, tiny or broad, regular or broken, symmetric or asymmetric? Look at other frames (selector is at upper right of window) as well!	Watch the green crosshair at ORGX ORGY, the green circle around it (lower INCLUDE_RESOLUTION_RANGE), the red UNTRUSTED_RECTANGLEs at the module borders, and the blue TRUSTED_REGION appear. The corresponding keyword=parameter lines in XDS.INP have the same colour code.
4. mask the shaded regions of the detector: at least the beamstop shadow, and the beamstop holder. There is a button Untrusted areas (...) towards the upper right with three tools: UNTRUSTED_ELLIPSE, UNTRUSTED_RECTANGLE, UNTRUSTED_QUADRILATERAL	This step is important, do not forget it! Reason is in the XDSGUI paper (reference at the end). More explanation of the tools is at XDSGUI#Frame . Wrongly placed UNTRUSTED areas can be removed by deleting their lines in XDS.INP .
5. click the XDS.INP tab and inspect its contents. For a first XDS run, typically you would leave everything at its defaults, except with broad reflections covering many pixels, set MINIMUM_NUMBER_OF_PIXELS_IN_A_SPOT to 6 instead of 3. In later XDS runs, modify parameters to optimize processing. JOB=XYCORR INIT COLSPOT IDXREF DEFPIX INTEGRATE CORRECT is only needed at the beginning i.e. for the first run.	The paradigm of manual processing is to repeat INTEGRATE and CORRECT with optimized parameters that are available after a completed XDS run. Please note that geometry parameters for INTEGRATE are in XPARM.XDS, not in XDS.INP !
6. run XDS click Save (unless greyed out) and Run XDS . The tasks of the JOB keyword are run one after the other. Thus watch the adjacent tabs to the right flash yellow as they are being filled with text and graphics. Inspect the left (text) and right (graphics) sides of these tabs. Once the CORRECT tab is filled, the XDS run has completed, and the reflection file XDS_ASCII.HKL is available. The goal is to make sure that spacegroup and resolution are correct or at least reasonable, and that the processing is optimal.	If XDS stops after the IDXREF step (with ERROR in IDXREF.LP), this most often happens because it could index less than 50% of the spots only. This may be due to ice rings that obviously cannot be indexed, or due to additional lattices. In most cases, this should at least prompt you to think about the possible reasons. To continue, set JOB=DEFPIX INTEGRATE CORRECT in the XDS.INP tab, Save and Run XDS.
7. inspection and analysis of diffraction patterns in the first run of XDS COLSPOT tab: is the "number of spots" evenly distributed among the frames? If not, it may be a sign of radiation damage or anisotropy, or plate-shaped crystal. Maybe see COLSPOT once IDXREF has run, inspect the predictions in Frame . Do they match the observed reflections? Visualize reflections in reciprocal space : in tools go to Further analyses and click the second button from the top. This will load the indexed reflections (yellow) and non-indexed ones (pink) as pseudo-PDB files into Coot. Inspect the lattices: is the "yellow lattice" convincing, and/or do the pink reflections form their own lattice(s), or are they just random?	IDXREF tab: are the "CLUSTER INDICES" of the difference vectors integer numbers, or close to integers? Are the cell parameters reasonable? Is the first POPULATION of the first SUBTREE close to 3000? Ice rings? See IDXREF.LP and IDXREF . INTEGRATE tab: are the curves smooth (good) or are there jumps (bad)? Try to think of reasons for jumps/spikes! Could it be the beamline flux or the crystal changing? Are their straight red and green lines all along the "Beam divergence" and "Mosaicity" plots? If not, too few strong reflections were found (inspect the "THREE-DIMENSIONAL PROFILE" output in the text part) and you should add DELPHI=20 in XDS.INP, or increase the parameter if the keyword already exists.
8. inspect the CORRECT tab: Numerical values and plots are only meaningful if the spacegroup is correct (screw axes don't matter for the statistics, though). If the spacegroup is unknown, the text part reports the spacegroup that XDS determines automatically. To do better, in tools / Further analyses click determine spacegroups with pointless. Output is in the terminal window! Since pointless is more elaborate and more believable (but not infallible), obtain the SPACE_GROUP_NUMBER and UNIT_CELL_CONSTANTS in XDS.INP from the pointless output (unless you know better, of course). Often there are two enantiomorphic spacegroups possible (Space group determination#Space group selected by XDS: ambiguous with respect to enantiomorph and screw axes, Space group determination#Table of space groups by Laue class and Bravais type) or even more than two; keep that in mind for the structure solution - the correct spacegroup is only found when the structure is solved and well refined.	The first look should go to the the first plot (I/sigma unmerged data). The red horizontal line is ISa - what is its value? The exact numerical value is in the text part, but the plot is enough for a good estimate. If the value is less than 5, something (likely indexing or spacegroup) is severely wrong. Good values are 20 and higher. See ISa. The blue line should be well below the red line, otherwise this is a sign of overexposure - the systematic error limits the quality of the data, not the random error.
9. if the spacegroup that CORRECT determined differs from the one that pointless determined, CORRECT should be re-run: change the JOB=... line to JOB=CORRECT, and go to step 6. If the spacegroup is correct, decide about the high-resolution cutoff: go to the section after STATISTICS OF SAVED DATA SET "XDS_ASCII.HKL" and inspect the table SUBSET OF INTENSITY DATA WITH SIGNAL/NOISE >= -3.0 . Put the upper resolution value into the INCLUDE_RESOLUTION_RANGE line in XDS.INP, and have JOB=CORRECT, go to step 6	A useful estimate of the high-resolution limit is the last resolution range that still gets a star "*" in the CC1/2 column; you may also consider the <I/sigma> column. FRIEDEL'_LAW=FALSE if the anomalous signal is so strong that you want to see the proper statistics (completeness ...), or whether scaling and merging should not take it into account (FRIEDEL'S_LAW=TRUE). Low resolution anomalous correlation above 60% is what I consider strong. Modify XDS.INP accordingly.
10. Second and further runs of XDS typically refine the upper resolution limit and optimize the geometric parameters. It is useful to compare, after each modification, the resulting CORRECT.LP with the best one previously available. This is achieved by items in tools / Saving and comparing good results and tools / Optimizing data quality	Thus, re-run INTEGRATE CORRECT or just CORRECT as needed (step 6). See Optimisation, XDSGUI#How to use XDSGUI

The XDSGUI paper (Brehm, Triviño, Krahn, Usón and Diederichs (2023) XDSGUI: a graphical user interface for XDS, SHELX and ARCIMBOLDO. J. Appl. Cryst. 56) is open access at https://doi.org/10.1107/S1600576723007057 .