4. Repeat steps 2 and 3 for the second and third components.
Figure 4b shows the third and eighth maps of the second
principal motion.
3.5 Partial Signal
Subtraction after
Multi-body Refinement
1. To perform focused classifications and/or refinements on part
of the complex after multi-body refinement, one can create
stacks of particles in which signal from the different bodies is
subtracted taking all their orientations into account. For example, to subtract all signal except the SF3b domain, type the
following commands from the project directory (also see Note
22):
mkdir Subtract
relion_flex_analyse --data MultiBody/job001/
run_data.star --model MultiBody/job001/run_model.star --bodies Example/4-bodies-tight-mask.
star --o Subtract/sf3b --subtract --keep_inside
Mask-and-Ref/SF3b_mask.mrc --ctf
On our computer, this calculation took 8 h.
2. The resulting STAR file with subtracted particles, “Subtract/
subtracted.star”, can then be used as input in a conventional masked 3D classification or refinement (see Note 23).
4 Notes
1. Multi-body refinement can be performed without
GPU-acceleration, but to keep computation times within reasonable limits you will need a multi-node CPU cluster, e.g.,
with more than 200 CPU cores.
2. Displaying the consensus map in a slice-viewer, e.g., for this
case by executing “relion_display --i Example/consensus_hal1_class001.mrc”, may help in determining
which parts of the complex are flexible.
3. The individual bodies are allowed to overlap. This may be
useful in characterizing the density at the interfaces between
different bodies (also see Note 19). In addition, by including
part of an adjacent larger body, overlapping bodies may help in
the refinement of relatively small, flexible domains.
4. RELION implements a “Mask creation” job-type that can be
used to low-pass filter an input map; binarize it at a specified
threshold; grow the binary mask by a specified number of
pixels; and add a soft, raised cosine-shaped, edge with a specified width. Besides this functionality, tools for generating 3D
masks are not provided in RELION. We find the “Volume
Eraser” tool in UCSF Chimera useful to manually edit 3D
maps; access it from from the “Tools” menu, under “Volume
154
Takanori Nakane and Sjors H. W. Scheres
Figure 4b shows the third and eighth maps of the second
principal motion.
3.5 Partial Signal
Subtraction after
Multi-body Refinement
1. To perform focused classifications and/or refinements on part
of the complex after multi-body refinement, one can create
stacks of particles in which signal from the different bodies is
subtracted taking all their orientations into account. For example, to subtract all signal except the SF3b domain, type the
following commands from the project directory (also see Note
22):
mkdir Subtract
relion_flex_analyse --data MultiBody/job001/
run_data.star --model MultiBody/job001/run_model.star --bodies Example/4-bodies-tight-mask.
star --o Subtract/sf3b --subtract --keep_inside
Mask-and-Ref/SF3b_mask.mrc --ctf
On our computer, this calculation took 8 h.
2. The resulting STAR file with subtracted particles, “Subtract/
subtracted.star”, can then be used as input in a conventional masked 3D classification or refinement (see Note 23).
4 Notes
1. Multi-body refinement can be performed without
GPU-acceleration, but to keep computation times within reasonable limits you will need a multi-node CPU cluster, e.g.,
with more than 200 CPU cores.
2. Displaying the consensus map in a slice-viewer, e.g., for this
case by executing “relion_display --i Example/consensus_hal1_class001.mrc”, may help in determining
which parts of the complex are flexible.
3. The individual bodies are allowed to overlap. This may be
useful in characterizing the density at the interfaces between
different bodies (also see Note 19). In addition, by including
part of an adjacent larger body, overlapping bodies may help in
the refinement of relatively small, flexible domains.
4. RELION implements a “Mask creation” job-type that can be
used to low-pass filter an input map; binarize it at a specified
threshold; grow the binary mask by a specified number of
pixels; and add a soft, raised cosine-shaped, edge with a specified width. Besides this functionality, tools for generating 3D
masks are not provided in RELION. We find the “Volume
Eraser” tool in UCSF Chimera useful to manually edit 3D
maps; access it from from the “Tools” menu, under “Volume
154
Takanori Nakane and Sjors H. W. Scheres