REL 2014-02-20 2014-02-21 2014-02-21 Youdong Mao Dana-Farber Cancer Institute Boston United States 02215 Joseph Sodroski Dana-Farber Cancer Institute Boston United States 02215 Mao Y Wang L Gu C Herschhorn A Desormeaux A Finzi A Xiang SH Sodroski JG 1 2015-10-27 REL

Restructured the entry directory to have "data" and ".private".

38.0 10.6019/EMPIAR-10008 EMDB EMD-5447 U115 /data picked particles - multiframe - unprocessed MRC MRC 124478 1 32 BIT FLOAT 256 0.74688 256 0.74688

U115dataset.mrcs: This is the particle set with the original pixel size. It is a subset of the larger dataset deposited as the *.dat SPIDER stack files (10007). U115D2dataset.mrcs: this was the particle stack used to reproduce the refinement in RELION starting from a Gaussian sphere. The dataset was two-fold binned from U115dataset.mrcs. We made minor modifications to the RELION version 1.1 source code. The purpose of these changes was two-fold: 1) to avoid pre-mature progression to finer angular sampling for images of lower contrast than those tested during the development of the original software, and 2) to solve a file resource interlocking issue that arises when a greater number of parallel nodes is employed. Three modifications were made in the file "ml_optimiser.cpp", as follows: (i) The function global_mutex.lock() was added at line 2220, global_mutex.unlock() at line 2242; (ii) 0.75 was changed to 0.05 at line 4684, allowing the finer angular sampling of lower contrast images in autoAdjustAngularSampling(); and (iii) Line 4656 was changed to "max_nr_iter_wo_resol_gain = (do_split_random_halves) ? 0 : 6" in autoAdjustAngularSampling(). These changes are not essential for successful refinement when one follows the author’s suggestion of starting from a 60-Å low-pass filtered model. However, we found the changes to be critical for successful refinement when a Gaussian sphere was used as a starting model.