Charting Complexity: Interactive Real-Time Visualizations of Large-Scale Networks and Embeddings with Helios-Web

Speaker: Filipi Nascimento Silva

Website: https://heliosweb.io

Abstract: Helios-Web is an open-source, web-based platform designed for real-time visualization and interactive analysis of massive-scale networks and high-dimensional embeddings. By using GPU-based rendering, modern web technologies (WebGL and WebAssembly), and parallel computing (via Web Workers), Helios-Web handles datasets with hundreds of thousands to millions of nodes in real-time. Its customizable force-directed layouts and advanced rendering features, including real-time density estimation and other advanced effects, help mitigate the ““hairball”” effect, ensuring even large, complex datasets are both visually coherent and easily interpretable. The versatility of the platform has been demonstrated in biological research (e.g., gene expression, single-cell RNA-Seq, metabolomics, and microbiome networks), social media analysis (e.g., Twitter, Mastodon, Bluesky), and the science of science (e.g., citation and collaboration networks). By supporting the visualization of KNN graphs and fuzzy simplicial set embedding (e.g., UMAP projections), Helios-Web also enables the exploration of high-dimensional embedding structures, making it an invaluable tool for uncovering multi-scale patterns in both traditional graph data and advanced machine learning pipelines. A key feature of Helios-Web under active development is its integration with Large Language Models (LLMs), enabling both text- and voice-based queries for network control and analysis directly through the visualization interface. This empowers users, from casual observers to domain specialists, to access sophisticated insights with minimal technical overhead. The modular architecture of the platform further allows it to function as a standalone tool or serve as a frontend (which can be loaded as library) for existing analytic workflows, making it easily adaptable to a wide range of environments and use cases. In this presentation, we will highlight both the existing and upcoming capabilities of Helios-Web by showcasing its application in three key areas: (a) the science of science, where it provides interactive 2D and 3D maps of scholarly data such as publications and authors; (b) social media analysis, through an integration with the OSoMe platform to visualize large-scale networks and embeddings; and (c) biological networks, where it serves to reveal intricate relationships among genes, metabolites, and microorganisms. Collectively, these functionalities represent a major step forward in our ability to visualize, understand, and interact with the complex systems that support large-scale data.