June 2019: 3-D display of local weighted CC_½ with option to apply a diffraction cut-off based on wCC_½, and 3-D display of the Kullback-Leibler divergence.

• The local per-reflection weighted CC_½ is now computed and displayed, and optionally may be used as the criterion for the diffraction cut-off instead of the default local mean I /σ(I).  There are two methods available to compute it: 1) a modification of the half-dataset random-selection method of Karplus & Diederichs, and 2) a modification of the 'σ-τ' method of Assmann, Brehm & Diederichs.  These methods make different assumptions so there are likely to be significant differences in the results obtained.  For each method the modification consists of using inverse-variance weighting; this is essential in the calculation of wCC_½ because significance tests on the standard Pearson correlation coefficient assume homoscedasticity (equal variances) of the intensity measurements, whereas in reality those measurements are strongly heteroscedastic.

  In the case of unmerged data input the values of the half-dataset mean intensities are added automatically to the input data and the random-selection method is used.  The weighted 'σ-τ' method is used for merged input data unless the user includes columns labelled 'IHALF1' and 'IHALF2' as the half-dataset mean intensities, in which case the first method is used.  The advantage of using wCC_½ over the mean I /σ(I) as the cut-off criterion becomes apparent when the estimates of σ(I) are inaccurate, as appears to be the case for serial data.  Note that the 'σ-τ' method also relies on the σ(I) values being accurate so should not be used if those values are questionable.  The inverse-variance weighting also uses the σ(I) values but since the weighting is relative they need not be on the same scale as the values of I_mean.

• The local mean Kullback-Leibler divergence D_KL has been added to the WebGL display.  D_KL is the posterior expectation of the log of the ratio of the posterior over the prior probability and is a measure of the amount of information gained from that intensity measurement over and above that gained from the default Wilson prior (so zero if no information was gained from the measurement, otherwise it's always positive).

December 2018:  WebGL Reciprocal lattice viewer enhancements.

• Made points depth-cued.
• Added 'Rock' & 'Roll' options, with speed & amplitude selection.  See help info for new bindings of 'q', 'r' and arrow keys.
• Added a new colour (9: pink) for 'red' points with I /σ(I) > 3 (complement function key binding changed to 'n').
• Added date stamp to Javascript sources to avoid caching problems.
• Added Get/Set orientation transfer facility (to provide identical orientations between views in different browser windows).
• Replaced 'Off/On' button text with clearer 'Hide/Show'.
• Added more function buttons to allow colour selection for each scene independently.
• Changed from frames to scrolling windows and added function buttons for keyboard shortcuts (mostly for the benefit of tablet & smartphone users!).

December 2018:  STARANISO citation added to FAQ.

August 2018:  3-D display of XDS profile-correlation CORR statistic added.

A fourth type of interactive colour-coded reciprocal-lattice display has been added when the input data are given as an unmerged XDS_ASCII.HKL file (not MTZ format).  The left-most scene then shows the value of the CORR statistic produced by XDS (item 11 in the file), i.e. for each reflection, the correlation coefficient between the observed reflection profile and the predicted profile that was used to integrate it.

The profile correlation plot gives a primary diagnostic of problems at the integration stage, as a mismatch of profiles could result from shortcomings in the sample (e.g. a cracked crystal) or in the profile estimation method (e.g. too coarse a partitioning of the detector surface).  As the profile correlation is given on a per-reflection basis, it is a more sensitive indication of such problems than the loss in <I /σ(I)> that it causes, since the latter quantity is a local average.

May 2018:  Updated privacy policy.

August 2017:  The reciprocal-lattice WebGL viewer now features the display of redundancies for unmerged data and selection of functions to be displayed.

March 2017:  The STARANISO server can now handle unmerged data (XDS ASCII or MTZ 'multi-record' format).

This development is in line with our philosophy of making as few prior assumptions as possible about the data (motto: \"Let the data speak!\").

Implicit assumptions made (incorrectly) by some software - namely that the anisotropic diffraction-limit surface is ellipsoidal with principal axes either parallel to the crystal axes or to the principal axes of the anisotropy tensor ellipsoid - were already relaxed in the previous version of the server, through a complete separation of the rôles of the anisotropic diffraction limit and of the anisotropy tensor.

With this new development of a protocol for anisotropy correction at the unmerged data stage, we go further.  By applying the anisotropic diffraction cut-off at an earlier stage in processing to the unmerged data, we try to ensure that weak measurements that can impact the quality of the final data by biasing the image scales and the error model can be eliminated early on in the process.

August 2016:  The 3-D reciprocal-lattice displays now use WebGL (JavaScript) instead of Java.

This requires that JavaScript is enabled in your browser, and no longer requires prior installation of a Java plug-in (nor indeed of any other additional software).

You are strongly advised to disable any Java plug-ins that were installed solely for the purpose of using previous versions of this server but are now no longer needed.