5. Determine the width, i.e., the standard deviation, of the Gaussian priors on the relative rotations and translations between
pairs of bodies. Rotational and translational searches of the
individual bodies will be performed locally, with a search
range of three times these standard deviations. Therefore,
these values should express the expectation by how much the
bodies move relative to each other in the data set. For the
spliceosome example, the fuzziness in the consensus structures
indicated that the core and the foot are relatively stable,
whereas the helicase, and especially the SF3b domains move
more. We thus used widths on the angular priors of 10
for the
core and foot, 15
for the helicase domain, and 20
for the
SF3b domain. Likewise, the widths on the translational priors
were set to two pixels for the core and foot, three pixels for the
helicase, and five pixels for the SF3b domain. For each project,
these values should be determined empirically, but the values
given here provide a useful starting point.
6. Combine the information from steps 1–5 into a single metadata STAR file. The file for the spliceosome example is stored
as” Example/4-bodies-tight-mask.star”.
3.2 Execute
Multi-body Refinement
1. Launch the RELION graphical user interface (GUI) from the
project directory. In case of the example data set, this is the
directory in which the Example, Mask-and-Ref and Micrographs directories are. If this is the first time you launch the
GUI in this directory, the program will ask whether you are
sure you want to launch the GUI. Type “y” to confirm.
2. On the left-hand, vertical list of job-types click on “3D multibody” .
3. On the “I/O” tab, set “Consensus refinement optimiser.star” to “Example/consensus_optimiser.star”.
This file contains all the information from the previously performed consensus refinement (see Note 6).
4. On the “I/O” tab, set “Body STAR file” to “Example/4bodies-tight-mask.star”. This is the STAR file that
defines the bodies, which was generated in Subheading 3.1,
step 6.
5. On the “I/O” tab, set “Reconstruct subtracted bodies”
to “yes”. This will make the program write out reconstructions
for each of the bodies where the density of the other bodies is
subtracted. Thereby, the density around each body should
become ever cleaner and also provide feedback on the subtraction quality, as the alignments of the bodies improve during the
subsequent iterations of multi-body refinement. When set to
“No”, fuzzy densities for the other bodies will be present
around the reconstructed density for each body.
150
Takanori Nakane and Sjors H. W. Scheres
pairs of bodies. Rotational and translational searches of the
individual bodies will be performed locally, with a search
range of three times these standard deviations. Therefore,
these values should express the expectation by how much the
bodies move relative to each other in the data set. For the
spliceosome example, the fuzziness in the consensus structures
indicated that the core and the foot are relatively stable,
whereas the helicase, and especially the SF3b domains move
more. We thus used widths on the angular priors of 10
for the
core and foot, 15
for the helicase domain, and 20
for the
SF3b domain. Likewise, the widths on the translational priors
were set to two pixels for the core and foot, three pixels for the
helicase, and five pixels for the SF3b domain. For each project,
these values should be determined empirically, but the values
given here provide a useful starting point.
6. Combine the information from steps 1–5 into a single metadata STAR file. The file for the spliceosome example is stored
as” Example/4-bodies-tight-mask.star”.
3.2 Execute
Multi-body Refinement
1. Launch the RELION graphical user interface (GUI) from the
project directory. In case of the example data set, this is the
directory in which the Example, Mask-and-Ref and Micrographs directories are. If this is the first time you launch the
GUI in this directory, the program will ask whether you are
sure you want to launch the GUI. Type “y” to confirm.
2. On the left-hand, vertical list of job-types click on “3D multibody” .
3. On the “I/O” tab, set “Consensus refinement optimiser.star” to “Example/consensus_optimiser.star”.
This file contains all the information from the previously performed consensus refinement (see Note 6).
4. On the “I/O” tab, set “Body STAR file” to “Example/4bodies-tight-mask.star”. This is the STAR file that
defines the bodies, which was generated in Subheading 3.1,
step 6.
5. On the “I/O” tab, set “Reconstruct subtracted bodies”
to “yes”. This will make the program write out reconstructions
for each of the bodies where the density of the other bodies is
subtracted. Thereby, the density around each body should
become ever cleaner and also provide feedback on the subtraction quality, as the alignments of the bodies improve during the
subsequent iterations of multi-body refinement. When set to
“No”, fuzzy densities for the other bodies will be present
around the reconstructed density for each body.
150
Takanori Nakane and Sjors H. W. Scheres