Release 3.0.6

This release has several improvements:

• RelionCor supports dose weighting of datasets at 100 kV.

• The inter-particle distance cutoff in the Laplacian-of-Gaussian (LoG) picker was improved.
  When LoG picker is enabled, minimum inter-particle distance in the "autopicking" tab is not used. This is by design. Instead, neighbouring particles within
  (detected diameter at the point + --LoG_diam_min) * --LoG_neighbour / 100
  are avoided. The default for --LoG_neighbour is 100 %. You can make this smaller or larger to pick more or less, respectively.

• Implemented --LoG_upper_threshold option.
  LoG picker tended to pick high contrast contamination like ice and ethane droplets. They have higher scores than real particles, so adjust default threshold is useless. Now we have --LoG_upper_threshold option to set an upper limit in scores. Useful values are typically 1.5 to 4. The default is 99999 (i.e. no limit in practice)

This release also contains the following bug fixes.

• PostProcess: fixed a bug where postprocess_masked.mrc is overwritten by half2_masked.mrc when --half_maps is provided (reported by @keitaroyam).
• Stability of histogram generation was improved.
• Prevented "too long arguments" errors in scratch (issue #467 reported by @EMBL-Structures).
• relion_star_handler --split: removed an assumption on the number of digits.
• AutoPicker: rejects 3D references for filaments (CCPEM https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1906&L=CCPEM&O=D&P=6094).

Release 3.0.5

This release fixes several problems:

• Class2D/Class3D/Refine3D crash when parallel disc I/O is no, using scratch and running on multiple nodes
• AutoPick writes a debug file
• Binary from GCC 8.0 behaves strangely
• The error message is confusing when the project directory is not writable
• The path to non-dose weighted sums is wrong in RelionCor's log files
• CPlot2D is not acknowledged in AUTHORS file

Release 3.0.4

This fixes crashes during initialisation of multibody refinement and build issues with Intel compilers.

• relion_flex_analyse preserves the original image names in the rlnImageOriginalName column
• Do dose-weighting?: Yes became the default in the GUI

Release 3.0.3

This update fixes an integer overflow bug in relion_refine(_mpi) at very fine angular samplings. Typically this manifested as segmentation faults at later steps of (multibody or normal) refinement.

This also fixes a bug in the MPI version of AutoPick, which printed a wrong total number of particles in the log (don't worry; the output STAR file was correct).

Release 3.0.2

This fixes a segmentation fault at the beginning of Refine3D. Thanks to @Fravadona for report.

Release 3.0.1

This update fixes a bug in MultiBody refinement (98171da) along with minor issues in the GUI.

2.1 Stable

Use "Slower, more exhaustive search" in CTFFIND 4.1 unless --fast_sea…

2.1.b1

v.2.1-beta-1

This fixes a critical bug affecting half-set (gold-standard) refinement.

Starting with version 2.0.4, each iteration of auto-refinement incorrectly
uses the same half-set reconstruction as for alignment of all data. The FSC
and resolution reported is however still based on comparing two half-set
reconstructions which contain completely separate image data. A
degree of reference-bias is however present. The effect is to report
a resolution higher than what should be under gold-standard conditions,
since, reconstructions are subject to a degree of over-fitting.

In order to claim a so-called "gold-standard" resolution, use any version
before and including 2.0.3, or version 2.1b1 (this release).

Stable

2.0.3

v2.0.3 (stable)