TL;DR
Researchers often struggle to visualize and share complex, large-scale biological trajectory data. This paper introduces "linus," a web-based tool that simplifies the exploration, sharing, and presentation of such data, making it accessible to a broader audience without requiring local software installations.
What Problem Does This Paper Address?
Biology is increasingly generating massive datasets of dynamic processes, represented as trajectories – think neural connections, cell movements, or animal paths. Visualizing these "hairball" datasets with traditional static methods is challenging due to overplotting and complexity. Existing tools often require specialized installations and are not ideal for collaborative exploration or easy sharing. The authors aim to bridge this gap by providing a user-friendly, web-based solution that allows researchers to interactively explore, share, and present these complex datasets in a concise, self-contained packet.
Methods
The core of linus is built on web technologies (JavaScript and WebGL) for the visualization part, ensuring broad accessibility. However, computationally intensive data processing is handled by a Python script. This script takes trajectory data from various formats (CSV, SVF, TGMM XML, biotracks) and enriches it with additional features like edge bundling (to highlight major pathways) and custom axes. The output is a "visualization packet" – a folder containing HTML, JavaScript, and related files – that can be opened in any web browser with WebGL support. Users can then interactively adjust visualization parameters (color maps, projections, clipping planes) directly within the browser. A key feature is the ability to seamlessly blend between different data states, such as original vs. edge-bundled views or 3D vs. 2D projections, facilitating deeper exploration.
Datasets
The paper showcases linus with examples from various biological applications, including:
- Zebrafish cell trajectories during gastrulation (from a previous work by the authors).
- Ant trails (2D+t).
- Blue whale and arctic tern GPS tracks (2D surface + t).
- Brain tractography (white matter tracts, 3D).
- Zebrafish explant cell tracks (3D + t).
The authors do not introduce a new benchmark dataset but rather demonstrate the versatility of linus across existing, real-world datasets.
Key Insights and Ideas
- Web-based accessibility is key: By leveraging web technologies, linus removes the barrier of local software installation, making complex visualizations accessible to a wider range of users, including those with limited programming experience.
- Interactive exploration is crucial for complex data: Static visualizations of large trajectory datasets are often overwhelming. Linus's interactive features, such as filtering, color mapping, and blending between different data states, allow users to actively dissect and understand the data.
- Sharing as a first-class citizen: The "visualization packet" concept, which bundles data and visualization into a single, shareable unit, significantly streamlines the process of communicating findings with collaborators.
- "Tours" for guided storytelling: The ability to create timed sequences of visualization actions and annotations allows researchers to craft guided narratives, making presentations and explanations more engaging and informative.
Key Results
- Linus can handle large datasets, with desktop computers capable of rendering over 10,000 trajectories at smooth framerates.
- Performance on mid-range smartphones is usable (around 10 fps), though less smooth.
- For VR applications (Oculus Go), performance is critical, and reducing the number of trajectories to around 1,000 is recommended for a comfortable user experience.
- The tool has been successfully used to explore and share complex cell tracking data from zebrafish embryos, facilitating collaboration and figure generation for publications.
Limitations and Future Work
- Performance variability: The authors acknowledge that performance varies across devices, recommending the creation of dedicated visualization packets optimized for specific output devices (e.g., desktop, mobile, VR).
- Future enhancements: The authors plan to incorporate more advanced preprocessing options like trajectory clustering and generic feature extraction. They also aim to extend the visualization capabilities to allow interactive data annotation, which could be used for downstream analysis like building trajectory classifiers. They also invite community contributions for new features and integrations.
Summary updated