scNodes: a correlation and processing toolkit for super-resolution fluorescence and electron microscopy

Correlative cryo-super resolution light and cryo-electron microscopy (SR-cryoCLEM) is an upcoming method that promises to enable accurately targeted, high-resolution in situ structure determination by combining the specificity of fluorescent labelling with the high resolution of cryoEM^1–4. As it is a novel and rapidly developing method, there is currently no dedicated open-source software available for data processing, and current implementations rely instead on numerous disjointed software packages for processing and visualization. Here, we present a solution: an open-source, standalone processing suite called ‘scNodes,’ that is specialized for (but not limited to) SR-cryoCLEM and can be used for super-resolution fluorescence data processing, image correlation, and fluorescence-guided particle picking for cryoEM. The project is available at github.com/bionanopatterning/scNodes

scNodes offers a simple user interface for super-resolution reconstruction using a node editor: an intuitive graphical interface that streamlines the entire workflow (Fig. 1a), from raw data to final reconstruction, by enabling on-the-fly adjustment of the processing setup, and by offering processing speeds more than 5× those achieved with ThunderSTORM⁵, a popular alternative (see Fig. S1). This fast real-time data processing is instrumental in allowing the user to develop an understanding of each processing step by observing its effect on the final outcome. This enables the user to customize the super-resolution reconstruction, leading to better results in a shorter time.

Figure 1. an overview of scNodes.

a) a view of the node editor (left) and the accompanying output display (right). The display shows the output of the active ‘PSF-fitting’ node, which is preceded in the processing pipeline by a particle detection, spatial filter, registration, and data import node. b) the correlation editor showing a super-resolution fluorescence image overlayed on a cryo-electron tomographic reconstruction.

scNodes also contains a correlation editor for correlation of light microscopy images with electron microscopy images and tomographic volumes. With this editor users can manually or semi-automatically perform rigid alignment, and correlate, visualize, and export fluorescence and electron microscopy datasets (Fig. 1b). While the editor was created to aid manual correlation of light to electron microscope images, a semi-automated procedure to align images of the same content type, based on TurboReg⁶, is also implemented. This procedure is highly suited for, e.g., cryo-electron tomography datasets, which typically contain multiple views of one sample at various magnifications: overview, search, exposure, and tomographic volume. The numerous alignment steps that are typically required in a CLEM experiment are facilitated by a number of internal plugins, including the aforementioned registration method as well as a hierarchical arrangement of the data in the scene that helps organise the entire correlation train in a straightforward way.

With the exception of cryoEM data processing, for which many excellent software packages are available (e.g., refs^7,8), the combined functionality of the node and correlation editors allows the user to perform all data processing steps, from raw fluorescence data to the final correlated super-resolution images, in a cohesive workflow. Once a correlated dataset has been prepared, scNodes can also be used for fluorescence-guided particle picking using built-in manual picking functionality.

The programme was designed from the ground up to be easily extensible. It is written in Python and has a modular structure that allows for customization through the use of standalone nodes and plugins. These can be added to the software with minimal effort - by following a simple template – and require little specialized knowledge to create. Plugins can be shared on discussions page of the GitHub repository, which also hosts a forum where users can ask for help, provide feedback, or request features.

Besides ease of use, processing speed was an important parameter in the design. scNodes makes use of parallel processing on the CPU to speed up operations such as image registration, filtering, and particle detection, as well as of GPU accelerated computing for point-spread function (PSF) fitting⁹ of fluorescence images and for drift correction.

Installation of the software is straightforward: releases can be downloaded from our GitHub repository and run without requiring setting up any local Python environment or dependencies. Alternative installation instructions, including instructions on how to set up a local development environment for the project, can be found in the manual (in the supplementary note).

An extensive user manual that contains tutorials and visual instructions on how to use the software is also provided. It includes: an introduction to the node editor (chapter 2), a tutorial on generating super-resolution reconstructions with either a PSF-fitting method or SOFI (chapter 2.4), a tutorial on creating custom nodes (chapter 2.6), or downloading external nodes (chapter 2.7), an introduction to the correlation editor (chapter 3), a tutorial on correlating super-resolution fluorescence and cryo-electron tomography data (chapter 3.5), a tutorial on writing and installing custom plugins for the correlation editor (chapter 3.6), fluorescence-guided particle picking (chapter 3.7), and various other sections that provide further information.

To ensure that scNodes can respond to the developments in the field and the need of users, we are continuously improving and extending the software and it is our hope that others who find the software useful may do so as well, and share their plugins on GitHub.