For detailed cluster analysis, the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) method [34] can be
used. However, to get optimal results, it needs optimization of its
parameters that can introduce a bias in the results. Recently
emerged Voronoi tessellation-based methods [9, 35–37] allow
not only for an unambiguous local density estimation and visualization of SMLM data but also for testing of datasets for clustering
via comparison of the Voronoi polygons built on the experimental
data with that of the reference distribution of points and for automated unbiased segmentation without parametrization [9, 37].
Some of these processing aspects can be found in packages for
microscope control, but in many cases, external tools have to be
used. There are several packages dedicated for processing of localization data, namely ThunderStorm [23] that includes postprocessing aspects such as image reconstructions in several modes,
filtering and combination of localizations, drift correction, and
estimation of colocalization. ViSP [38] is a convenient tool for
3D visualization and basic segmentation. Lama [39] provides
tools for quality control, resolution estimation, cluster analysis
with Ripley’s functions, DBSCAN and Ordering Points To Identify
the Clustering Structure (OPTICS) [40] algorithms, coordinatebased colocalization, and estimation of stoichiometry of molecular
complexes [12]. MIiSR [41] is focused on colocalization estimation and cluster analysis using the Ripley’s, DBSCAN, and OPTICS
methods. PALMsiever [42] allows for the combination of localizations, drift correction, DBSCAN cluster analysis, and different
rendering possibilities. SharpViSu [29] offers comprehensive tools
for the correction of chromatic aberrations and drift, selection and
filtering of localizations, different visualization modes, resolution
estimation and, through ClusterViSu [9], cluster analysis with the
Ripley’s function method and Voronoi-based automated segmentation. Our latest development, 3DClusterViSu, allows to perform
Voronoi-diagram-based segmentation of 3D data [37]. Clearly, if it
can be done for a given experiment, 3D analysis becomes the
method of choice nowadays as compared to 2D analysis because it
avoids artefacts from superposed structures (see details in Supplementary Material in [37]), and it provides direct 3D information
about the molecular organization in the cell, thus also facilitating
the integration with molecular and atomic structures and opening
the path toward cellular structural biology [1].
4 Notes
1. Absence of any blocking reagents and blocking step is possible
with highly specific antibodies. If necessary, BSA 1% or a mixture of decomplemented NGS (5%) and FCS (1%) can be used
before step 6.
Practical Aspects of Super-Resolution Imaging and Segmentation of. . .
283
used. However, to get optimal results, it needs optimization of its
parameters that can introduce a bias in the results. Recently
emerged Voronoi tessellation-based methods [9, 35–37] allow
not only for an unambiguous local density estimation and visualization of SMLM data but also for testing of datasets for clustering
via comparison of the Voronoi polygons built on the experimental
data with that of the reference distribution of points and for automated unbiased segmentation without parametrization [9, 37].
Some of these processing aspects can be found in packages for
microscope control, but in many cases, external tools have to be
used. There are several packages dedicated for processing of localization data, namely ThunderStorm [23] that includes postprocessing aspects such as image reconstructions in several modes,
filtering and combination of localizations, drift correction, and
estimation of colocalization. ViSP [38] is a convenient tool for
3D visualization and basic segmentation. Lama [39] provides
tools for quality control, resolution estimation, cluster analysis
with Ripley’s functions, DBSCAN and Ordering Points To Identify
the Clustering Structure (OPTICS) [40] algorithms, coordinatebased colocalization, and estimation of stoichiometry of molecular
complexes [12]. MIiSR [41] is focused on colocalization estimation and cluster analysis using the Ripley’s, DBSCAN, and OPTICS
methods. PALMsiever [42] allows for the combination of localizations, drift correction, DBSCAN cluster analysis, and different
rendering possibilities. SharpViSu [29] offers comprehensive tools
for the correction of chromatic aberrations and drift, selection and
filtering of localizations, different visualization modes, resolution
estimation and, through ClusterViSu [9], cluster analysis with the
Ripley’s function method and Voronoi-based automated segmentation. Our latest development, 3DClusterViSu, allows to perform
Voronoi-diagram-based segmentation of 3D data [37]. Clearly, if it
can be done for a given experiment, 3D analysis becomes the
method of choice nowadays as compared to 2D analysis because it
avoids artefacts from superposed structures (see details in Supplementary Material in [37]), and it provides direct 3D information
about the molecular organization in the cell, thus also facilitating
the integration with molecular and atomic structures and opening
the path toward cellular structural biology [1].
4 Notes
1. Absence of any blocking reagents and blocking step is possible
with highly specific antibodies. If necessary, BSA 1% or a mixture of decomplemented NGS (5%) and FCS (1%) can be used
before step 6.
Practical Aspects of Super-Resolution Imaging and Segmentation of. . .
283