[3dem] RELION release

Sjors Scheres scheres at mrc-lmb.cam.ac.uk
Mon Nov 28 06:57:29 PST 2011

Dear colleagues,

I am pleased to announce the release of a new program for 
single-particle refinement, called RELION (for REgularised LIkelihood 
OptimisatioN). RELION implements a novel, Bayesian approach to 
single-particle refinement, which is explained in detail in the 
following paper: http://dx.doi.org/10.1016/j.jmb.2011.11.010  (JMB, in 
press). RELION may be used to calculate reference-free 2D class 
averages, to classify structurally heterogeneous data sets in 3D, and to 
refine structurally homogeneous data sets to high resolution.

In brief, RELION iteratively learns the optimal linear (or Wiener) 
filter for each reconstruction from the data themselves. In this way, 
the reconstruction(s) are simultaneously low-pass filtered and corrected 
for CTF effects without the need for user supervision. Thereby, 
subjective decisions related to arbitrary low-pass filters and/or 
heuristic approximations to the 3D Wiener filter are avoided. The 
resulting algorithm is effective in reducing overfitting, obtaining 
objective resolution estimates, and detecting small classes in 
structurally heterogeneous data sets.

RELION is open-source and may be downloaded for free from: 
I hope this new program will contribute positively to your research, and 
would appreciate your feedback on it.

Best regards,

Sjors Scheres
MRC Laboratory of Molecular Biology
Hills Road, Cambridge CB2 0QH, U.K.
tel: +44 (0)1223 402217

More information about the 3dem mailing list