The VRNetzer platform enables interactive network analysis in Virtual Reality

Network theory provides a robust framework for analyzing diverse data, finding numerous applications in biology and medicine for untangling complexities across biological organizational levels. The visual nature of networks allows for intuitive interpretation, enabling the identification of local and global patterns that can be further analyzed computationally. For example, in molecular interaction networks, highly connected hubs often represent genes with significant roles in healthy and disease states. Dense clusters often indicate specific biological functions, while disease-associated processes are characterized by distinct connection patterns. Interactive visualizations are crucial for exploring functional annotations of genes. However, current methods are limited by the size of networks that can be effectively visualized and interactively explored on conventional computer screens; large networks often become unintelligible "hairballs". This limitation necessitates the selection of subnetworks, but determining the most relevant subnetwork is challenging, particularly as biological networks exhibit the small-world property, meaning large parts of the network are reachable within a few steps. This paper addresses these challenges by utilizing immersive Virtual Reality (VR) technology. The immersive nature of VR allows for a sense of depth facilitating the resolution of connections obscured in 2D or non-immersive 3D representations. This allows for visualizing and interactively exploring significantly larger networks, enabling the seamless inspection of local and global network structures. The authors present network layouts specifically tailored for this immersive environment and demonstrate their application in investigating gene variants associated with severe genetic diseases.

The paper cites numerous existing software solutions for network visualization, highlighting their strengths and limitations. These range from standalone software like Cytoscape and Gephi, which offer interactive visualization but struggle with very large networks, to web-based tools like OmicsNet and GraphVis, which also have limitations in handling the scale and interactivity needed for genome-scale network analysis. The authors also discuss specialized tools for gene annotation and prioritization such as ToppGene, HGC, and STRING, highlighting their strengths in terms of analysis capabilities but emphasizing the absence of interactive visualization, limitations on customizability, and dataset integration. This review sets the stage for the presentation of VRNetzer as a novel solution aiming to address the shortcomings of existing tools.

VRNetzer is a modular platform consisting of five key modules: (1) a data module using a MySQL database for storing and organizing data; (2) an analytics module using Python for data access and user-defined analyses; (3) a VR module implemented in Unreal Engine for immersive 3D visualization; (4) a user interface (UI) module acting as a communication layer using standard web design libraries; and (5) a web module providing a browser-based frontend for tasks better suited to conventional computers. The VR module supports various navigation modes, allowing for free movement, rotation, translation, and scaling within the network. It features functionalities for annotating and analyzing networks at different scales, from individual nodes to clusters and global patterns. Custom layouts are designed to leverage the 3D space, facilitating the spatial organization of nodes based on their functional or structural characteristics. The user interface uses HTC Vive controllers, enabling natural 3D gestures for manipulating the network. It also includes 2D panel interfaces for more programmatic tasks, providing a control panel for network exploration and an inspection panel for detailed information. The modular design allows for easy extensibility by using standard web development libraries. A web-based frontend provides a seamless transition between VR and standard workflows. The data module is populated with diverse biological datasets including protein-protein interactions, gene annotations (GO terms, disease associations), pathway annotations, and gene expression data. The analytics module allows implementing custom data analysis methods in Python. A five-step process for gene variant prioritization is presented, integrating data preparation (web module), VR exploration, disease neighborhood identification using random walk with restart, detailed variant inspection, and post-processing. The paper includes detailed information on the implementation of each module, highlighting the use of specific technologies and libraries.

VRNetzer successfully enables interactive exploration of genome-scale molecular networks (e.g., the human interactome with ~16,000 proteins and ~300,000 interactions) in VR. Its immersive nature significantly improves the ability to resolve connections and visualize both local and global network structures, overcoming the limitations of conventional 2D visualizations. The platform’s modular design ensures flexibility, facilitating customization of data visualization, analysis, and data input. The dynamic transition functionality allows users to seamlessly explore different functional contexts, providing a powerful tool for investigating gene functions and their roles in disease. The application to gene variant prioritization in a rare disease case study demonstrates the platform’s effectiveness in identifying disease-causing genes by integrating expert knowledge, computational analysis, and visual network exploration. The five-step process, integrating web and VR functionalities, efficiently guided the analysis, leading to the identification of DOCK2 as the most likely candidate gene. The visualization of the network in 3D allowed for easier identification of relevant subnetworks and relationships between genes, greatly enhancing the understanding of the biological mechanisms involved in the disease. The performance of the platform is robust, with analyses and database calls completed in under a second, even on entry-level hardware. The open-source availability of VRNetzer ensures broader community access and further development.

VRNetzer addresses a critical need in the analysis of large-scale biological networks. Its immersive VR environment provides a uniquely intuitive and effective way to explore complex relationships, integrating computational tools and expert interpretation. The success in identifying a disease-causing gene in a rare disease case study demonstrates the platform’s value for translating complex data into biologically relevant insights. The modular design allows for adaptation to various datasets and research questions, extending its applicability beyond molecular networks. The open-source nature of the platform fosters collaboration and further development, contributing to the advancement of data exploration methods in diverse scientific fields. The findings highlight the potential of VR and related technologies for enhancing data exploration and analysis in a way that leverages both human cognitive capabilities and advanced computational methods.

VRNetzer offers a significant advance in interactive network analysis by leveraging the immersive capabilities of VR. Its modular design, open-source availability, and successful application in rare disease gene identification demonstrate its potential to transform how scientists explore complex datasets. Future work could explore incorporating additional data types, enhancing the user interface, and expanding analytical functionalities. The platform represents a promising step towards bridging the gap between complex data and intuitive understanding.

The current version of VRNetzer requires VR hardware, which may limit accessibility. Although the platform can run on various hardware configurations, high-end hardware offers superior performance and visual experience. While the platform is highly customizable, some technical expertise may be required for advanced modifications and integration with new data sources. Future work should focus on enhancing user-friendliness and reducing the technical barriers to entry for non-expert users.