coefficients between right- and left-handed template structures. Theses
approaches are fast and do not need much user input, but high-scoring false
positives are retained and have to be sorted out later by classification approaches. A third approach is to (C) manually inspect peaks in the context of the
tomogram to assign their identity (Fig. 9.1a). This typically yields a very pure
set of subtomograms with few false positives, but is quite laborious.
(A) Display a histogram of correlation coefficients for all extracted peaks
(column 1 in TOM/av3 motiflists). For data with high quality, a Gaussian
distribution of high correlation coefficients originating from ‘true positives’ will be visible and mostly separated from an increasing number of
low correlation coefficients originating from ‘false positives’ (Fig. 9.1b).
A cross-correlation threshold placed on the left-hand tail of the Gaussian
distribution discriminates roughly between true and false positives.
(B) If template matching has been performed for a right-handed and a
left-handed template structure, selection of peaks can be based on a
comparison of their correlation coefficients. Plot the correlation coefficients for the right and left-handed templates into the same figure. For
true positive peaks, cross-correlation values for the right-handed template should be higher than for the left-handed template, while they
should be similar for false positive peaks. Thus, the curve for the
right-handed template should start clearly separated from the curve for
the left-handed template in the region of the highest-scoring peaks, but
both curves should converge for lower cross-correlation values, indicating the number of true positive peaks detected in the tomogram
(Fig. 9.1c).
(C) Peaks can be inspected visually in the context of the tomogram (tom_chooser.m) and the motiflist can be filtered according to the assigned
class labels (Fig. 9.1a).
9.8.2 Extraction and Alignment of Subtomograms
This section introduces how to extract and align subtomograms iteratively to
increase resolution and signal-to-noise ratio. The main steps are (I) reconstruction
of subtomograms at the selected coordinates and (II) running iterative subtomogram
alignment using different approaches.
(I) After localization of particles, subtomograms are individually reconstructed
from the unbinned aligned weighted projections at the selected coordinates
provided in lines 8–10 (x,y,z) in TOM/av3 motiflists (av3_fastrecparticles.m).
Note that the particle coordinates must be provided with respect to the center
9 Structural Biology in Situ Using Cryo-Electron …
253
approaches are fast and do not need much user input, but high-scoring false
positives are retained and have to be sorted out later by classification approaches. A third approach is to (C) manually inspect peaks in the context of the
tomogram to assign their identity (Fig. 9.1a). This typically yields a very pure
set of subtomograms with few false positives, but is quite laborious.
(A) Display a histogram of correlation coefficients for all extracted peaks
(column 1 in TOM/av3 motiflists). For data with high quality, a Gaussian
distribution of high correlation coefficients originating from ‘true positives’ will be visible and mostly separated from an increasing number of
low correlation coefficients originating from ‘false positives’ (Fig. 9.1b).
A cross-correlation threshold placed on the left-hand tail of the Gaussian
distribution discriminates roughly between true and false positives.
(B) If template matching has been performed for a right-handed and a
left-handed template structure, selection of peaks can be based on a
comparison of their correlation coefficients. Plot the correlation coefficients for the right and left-handed templates into the same figure. For
true positive peaks, cross-correlation values for the right-handed template should be higher than for the left-handed template, while they
should be similar for false positive peaks. Thus, the curve for the
right-handed template should start clearly separated from the curve for
the left-handed template in the region of the highest-scoring peaks, but
both curves should converge for lower cross-correlation values, indicating the number of true positive peaks detected in the tomogram
(Fig. 9.1c).
(C) Peaks can be inspected visually in the context of the tomogram (tom_chooser.m) and the motiflist can be filtered according to the assigned
class labels (Fig. 9.1a).
9.8.2 Extraction and Alignment of Subtomograms
This section introduces how to extract and align subtomograms iteratively to
increase resolution and signal-to-noise ratio. The main steps are (I) reconstruction
of subtomograms at the selected coordinates and (II) running iterative subtomogram
alignment using different approaches.
(I) After localization of particles, subtomograms are individually reconstructed
from the unbinned aligned weighted projections at the selected coordinates
provided in lines 8–10 (x,y,z) in TOM/av3 motiflists (av3_fastrecparticles.m).
Note that the particle coordinates must be provided with respect to the center
9 Structural Biology in Situ Using Cryo-Electron …
253