Picasso-server: a community-based, open-source processing framework for super-resolution data

Super-resolution microscopy has become increasingly robust and performant over the last decade. Here, we report a server infrastructure that combines robust community-tested open-source processing algorithms with a workflow management system and database to enable production-grade processing of high-content super-resolution data.

(i.e., with nearest neighbor analysis and the NeNA 9 value), and kinetics. While these properties are often manually calculated and inspected when examining a reconstructed image, Picasso-server chains them to the localization part in the pipeline (Fig. 1b). This allows having multiple meaningful metrics to characterize an experiment and is directly inspired by similar approaches in proteomics.
Second, Picasso-server provides a web interface to interact with the local database (Fig. 1c). Here, we rely on Streamlit, which has rapidly become one of the most popular web-hosting frameworks for Python and which we have successfully used for proteomics software already 10 . A history tab allows exploring the summary statistics and inspecting instrument performance. Here, database entries can be filtered by date and keywords and grouped. Potential deviations in experiment performance (e.g., decreasing photon numbers) become readily visible, facilitating the technical assessment and effectively scheduling maintenance cycles, identifying optimal imaging conditions or localization settings. (Fig. 1c). In an additional compare tab, users can directly compare multiple experiments using the processed molecule lists. This allows advanced comparisons such as investigating the localizations per time or comparing distributions. The compare tab is meant to simplify troubleshooting individual experiments and, e.g., identifying differences compared to a reference run. To further facilitate bookkeeping of recorded experiments, Picasso-server has a Preview tab that allows rendering of the singlemolecule localizations directly within the browser.
Third, Picasso-server incorporates a file watcher that can automatically detect new files and process them with preselected settings. This includes functionality to call other Picasso functions or completely custom commands like Notifications to build tailored automation pipelines. The intended use case of the watcher relates to a lab environment with multiple instruments that continuously produce data and a central processing computer to which files are transferred (Fig. 1d). Users can therefore monitor existing experiments directly via the web server, making the access, e.g., via Remote Desktop, obsolete. Here, they can conveniently monitor the files that are currently being processed and assure performance.
Picasso-server extends the community-tested algorithms of Picasso and makes them accessible from a server interface. By connecting summary statistics and additionally quality metrics as part of a default analysis workflow, users can effectively conduct and monitor large-scale experiments. Ultimately, we hope that Picasso-server enables the next generation of high-throughput super-resolution studies. Further extensions to Picasso-server could include notification systems, automated performance warnings, or tailored recommendations on how to improve imaging quality. Like Picasso, Picasso-server is readily available as open-source with a one-click installer and a permissive MIT license. Fig. 1 Overview of the Picasso-server functionality and workflow. a Picasso "Render"-Interface displaying super-resolution data. b Server-Functionality: Metadata, summary statistics, and metrics derived from postprocessing algorithms are added to the database after processing. c Left side: Web interface for Picasso-server. The database and connected raw data can be explored via multiple tabs; here, shown is the History tab. Right side: Example plots for visualizing instrument performance in the history tab. The upper panel shows Boxplots of the NeNA estimate for localization precision per day; the lower panel shows the trendline in the number of collected Photons over time. d Example multi-user, multi-microscope setup. Data can be collected from multiple microscopes and is transferred to a processing PC, where files are automatically processed and indexed in the database. Multiple users can access Picassoserver to track instrument performance and progress.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Sample data used in the accompanying demonstration video was taken from the Single Molecule Localization Microscopy Challenge 2016, https://srm.epfl.ch.

Code availability
The source code can be found at https://github.com/jungmannlab/picasso, which contains current releases and a link to the documentation. The Picasso package is also available in the Python Package Index. A standalone version for Windows is available on the release page. We additionally provide a Dockerfile for cross-platform support or deployment in a cloud environment. Detailed links to the individual code parts and installation instructions for Picasso-server are included in Supplementary Note 1. The documentation is additionally available as Supplementary Note 2.