ViSP: representing single-particle localizations in three dimensions

To the Editor: Recent developments in three-dimensional (3D) methods of localizing stochastically activated fluorescent probes have advanced our understanding of the organization of biological systems at the nanoscale level1. Techniques such as photoactivated localization microscopy (PALM) and stochastic optical reconstruction microscopy (STORM) make it possible to probe the morphology of subcellular structures and detail their dynamic behavior in three dimensions (with spatial resolutions as small as 10 nm (ref. 2)). Although powerful programs for localizing 3D single-particle detections are publicly available (such as QuickPALM3, rapidSTORM4 and the μManager localization microscopy plug-in), user-friendly tools for representing generated localization data in a biologically relevant manner remain in great demand in the super-resolution microscopy community. In this regard, among the greatest challenges include the efficient handling of millions of individual localizations; the description of localization precisions in three dimensions; and the incorporation of intuitive depth cues, multichannel compatibilities and 3D quantitative features. With no obvious solutions in either the commercial or public software domains, we developed ViSP: an interactive, freely available, cross-platform 3D localization representation tool (Supplementary Software). Individual localizations in ViSP are visually represented by their intensities and localization precisions in three dimensions (as determined by single-particle localization software). Each localization can be rendered effectively as an anisotropic 3D Gaussian profile, with s.d. corresponding to the respective localization precisions in the lateral and axial dimensions (Fig. 1a and Supplementary Video 1). Despite their individualized visual treatment, data sets consisting of millions of localizations are displayed rapidly and interactively. Because the representation of point-cloud data as Gaussian profiles incurs a loss in resolution5, ViSP can additionally render 3D localizations in the form of scatter plots, octree-based histograms or, as discussed below, density plots. Color maps can be applied to axial depth, intensity or frame number as well as the relative density of individual localizations (Fig. 1b). Density plotting is a practical feature for studying morphologies, molecular clustering and accumulation of molecules at a specific site. By setting a threshold, users can filter out low-density points in a preset manner, a feature that is critical for microscopy techniques and samples that are susceptible to a high number of sparse artifactual detections. A powerful method for interpreting high-density groupings of 3D localizations as they relate to potential biological structures is through surface rendering with visually intuitive depth cues such as lighting, shading, fog (aerial) and perspective effects6. ViSP accomplishes this by rendering closed surfaces around localizations on the Figure 1 | Visualization of 3D single-particle localizations with ViSP. (a) Rotation sequence of a cropped region from a 3D STORM reconstruction of mammalian mitochondria in which the poorer localization precision in the z axis is revealed. (b) 3D density plot of a PALM-reconstructed actin cytoskeleton in a HeLa cell. Bottom, spatial histogram of the localizations inside the green cylinder piercing two of the actin filaments. (c) Surface reconstruction and cluster segmentation of a mammalian mitochondrial network as in a. Left, localizations color coded to depth (z). Right, surface rendering and cluster segmentation of the same localizations. (d) Multichannel 3D PALM/STORM localization overlay of a centrosome complex: Centrin1 (red) and Cep 164 (blue). Left, density plot overlay of the localizations from both channels; right, overlay of the corresponding surfaces. Dimensions are in nanometers and densities are shown in arbitrary units (au). a