[buster-discuss] Serious bug in CCP4 TRUNCATE program

[proc-develop]

Dear BUSTER users,

a serious bug in the CCP4 [1] program TRUNCATE [2] has come to our
attention, which we have just reported to the CCP4 team. This problem
affects all computations involving conversion from intensities to
amplitudes, resulting in wrong amplitudes/sigmas. It was introduced on
"Mon 2017-05-22 17:19:55 +0100" (via an only partially correct
modification by CCP4) and possibly made it into CCP4 update version
7.0.039 onwards [3], i.e. it is still present in the current CCP4
version 7.0.056.

Please note that the data autoPROC [4] provides via our own STARANISO
[5] program (staraniso_alldata-unique.mtz) is unaffected (since it
handles that procedure internally itself in the correct way). We are
still investigating the impact this bug could have on refinement
results when using the problematic data (truncate-unique.mtz), but are
aware of a number of deposited PDB entries during the last year
affected by this [6].

The recommended solution for autoPROC users is to use the STARANISO
processed data. The next autoPROC release will contain additional
checks and workarounds for this problem until a fix has been released
by the CCP4 team.

Kind regards

Clemens, Claus, Ian, Andrew & Gerard (for autoPROC developer)

[1] http://www.ccp4.ac.uk/
[2] http://www.ccp4.ac.uk/html/truncate.html
[3] http://www.ccp4.ac.uk/updates/
[4] http://www.globalphasing.com/autoproc/
[5] http://staraniso.globalphasing.org/
[6] https://www.globalphasing.com/autoproc/wiki/index.cgi?IssuesPage20171219