Odon: An ultra-fast viewer for spatial proteomics

Odon is a high-performance, native Rust desktop viewer designed for the rapid, interactive visualization of large-scale spatial proteomics and transcriptomics datasets, offering superior loading speeds and rendering capabilities compared to existing tools like QuPath and Napari.

The Gist

Imagine you are trying to look at a massive, incredibly detailed map of a city, but instead of streets and buildings, the map is made of millions of tiny cells inside a human tissue sample. This is what scientists do with spatial proteomics: they take pictures of tissues to see how different proteins interact, often capturing hundreds of images of the same slice of tissue to get a 3D-like understanding of how a tumor or healthy tissue works.

The problem? These maps are huge. We are talking about digital files so large (sometimes 32 gigabytes or more) that they are like trying to open a library of books all at once on a standard laptop. Usually, to view these files without your computer freezing, you need a super-expensive, specialized computer the size of a small refrigerator. If you have a lab full of researchers, they all have to wait in line to use that one giant machine.

Enter Odon.

What is Odon?

Think of Odon as a super-powered, lightweight magnifying glass that runs on a regular laptop. It was built by researchers at UCL to let scientists zoom in and out of these massive tissue maps instantly, without needing a supercomputer.

Here is how it works, using some simple analogies:

1. The "Dragonfly" Speed

The tool is named Odon after the dragonfly (Odonata). Dragonflies are famous for their incredible speed and ability to see flickering light that other animals miss. Similarly, Odon is designed to be ultra-fast.

• The Old Way: Imagine trying to load a 32GB image on a standard viewer like QuPath or Napari. It's like trying to drink a swimming pool through a soda straw. It takes 10 to 35 seconds just to get the first sip, and if you try to zoom in, the straw might snap (the computer crashes).
• The Odon Way: Odon drinks that same swimming pool in less than one second. It's like having a firehose instead of a straw.

2. The "Smart Delivery" System

Why is it so fast? It uses two main tricks:

• The "Lego" Format (OME-Zarr): Instead of trying to load the entire massive image file at once (like trying to carry a whole brick wall), Odon breaks the image into tiny "Lego bricks." It only loads the specific bricks you are looking at right now. If you zoom out, it swaps them for bigger, lower-detail bricks. If you zoom in, it swaps them for high-detail bricks. This happens so fast you don't even notice.
• The "Rust" Engine: Most scientific software is built with languages like Python or Java, which are great but can be a bit slow and heavy (like driving a bus). Odon is built with Rust, a programming language known for being incredibly efficient and lightweight (like driving a Formula 1 race car). This allows it to run smoothly on a standard laptop rather than needing a massive server.

3. The "Mosaic" View

One of the coolest features is the Mosaic Mode.

• Imagine you have a tissue sample with 100 different tiny cores (like a cookie cutter taking 100 bites out of a cookie). Usually, you'd have to open 100 different windows to look at them.
• Odon lets you see all 100 of those tiny samples on one single screen at the same time. It's like having a panoramic window where you can see the whole neighborhood at once, but if you see something interesting in one house, you can instantly zoom in to see the details of that specific room.

4. Why Does This Matter?

Currently, if a lab wants to check if their images are blurry or if the staining is bad, they often have to wait for a powerful machine to become free.

• Before Odon: "I need to check my data. I'll have to wait 2 hours for the big computer to be free."
• With Odon: "I need to check my data." Click. It's done in a second on their regular laptop.

The Bottom Line

Odon is like giving every scientist a Ferrari instead of a tractor. It removes the bottleneck of needing expensive, specialized hardware. It allows researchers to look at massive, complex biological data instantly, spot errors quickly, and analyze hundreds of samples at once, all from a standard laptop. This means science can move faster, and more people can work on it at the same time.

Technical Summary

1. Problem Statement

The field of spatial proteomics and transcriptomics is generating massive, high-dimensional imaging datasets (often spanning entire slides and reaching terabytes in size for cohorts). These datasets pose significant challenges for visualization and quality control (QC):

• Hardware Bottlenecks: Current workflows often require specialized, high-end workstations (e.g., 12-core processors, 256 GB+ RAM, 16 TB SSD) to visualize data efficiently. This creates a bottleneck where only one researcher can analyze data at a time, hindering productivity in large labs.
• Software Limitations: Existing popular viewers (e.g., QuPath, Napari), often built on Java or Python, struggle with performance. They exhibit slow loading times for large whole-slide images, crash when zooming into high-resolution data, and fail to render dense segmentation data (typically plateauing around 50,000 objects).
• QC Needs: Rapid visual inspection is critical for detecting staining artifacts (e.g., protein aggregates, non-specific staining), but current tools are too slow for efficient, interactive exploration at the cohort scale.

2. Methodology

The authors developed Odon, a native desktop viewer designed to overcome these limitations through two core design principles:

• Programming Language (Rust): Odon is built using Rust, a systems programming language known for memory safety and high performance. This choice leverages Rust's superior execution speed and multi-threading capabilities compared to the Python and Java backends used by competitors like Napari and QuPath.
• Data Format (OME-Zarr): The viewer is optimized for the OME-Zarr format, a cloud-native, chunked imaging format.
  • Lazy Loading: Instead of downloading entire datasets, Odon uses viewport-driven tile loading, fetching only the necessary data chunks from local storage or remote sources (HTTP/S3).
  • Efficiency: OME-Zarr reduces reliance on single file handles, improving chunk retrieval times locally and remotely compared to traditional TIFF formats.
• Rendering Architecture:
  • GPU-Based Compositing: Odon utilizes a GPU-accelerated pipeline for rendering, enabling smooth interaction with massive datasets.
  • Unified Layer System: It supports a unified system for overlaying images, labels, and masks, allowing real-time interrogation of single-cell data.
• Supported Formats: Primary support for OME-Zarr, GeoJSON, and GeoParquet. Secondary support includes SpatialData, Xenium containers, and TIFF.

3. Key Contributions

• Ultra-Fast Performance: Odon drastically reduces image loading times and enables interaction with datasets that previously caused system crashes in other tools.
• Scalable Segmentation Rendering: It can smoothly render and interact with over 1,000,000 segmented cells without performance degradation, far exceeding the ~50,000 object limit of traditional viewers.
• Cohort-Level Visualization (Mosaic Mode): A dedicated "mosaic mode" allows users to view hundreds of Regions of Interest (ROIs) simultaneously on a single canvas. This is driven by project workspaces or sample sheets, enabling rapid comparison of marker expression across entire cohorts or Tissue Microarrays (TMAs).
• Integrated Visual Analytics: The tool includes live thresholding, cell selection, and on-canvas histograms/scatter plots, allowing users to perform analysis directly within the viewer without exporting data.
• Accessibility: By running efficiently on standard consumer laptops (e.g., MacBook Pro), Odon democratizes access to spatial proteomics analysis, removing the need for specialized, expensive workstations.

4. Results

The authors benchmarked Odon against QuPath (v0.7.0) and Napari (v0.6.6) using a 32 GB, 36-plex whole-slide OME-Zarr image (TNP_pilot_cycif dataset) on a 2024 MacBook Pro (M4 chip, 16 GB RAM).

• Loading Speed:
  • Odon: Loaded the full image in < 1 second.
  • QuPath: Took 10.14 seconds.
  • Napari: Took 35 seconds.
• Stability: While QuPath caused a system crash when the user attempted to zoom into the high-resolution image, Odon maintained stability.
• Segmentation Handling: Odon successfully rendered and interacted with >1,000,000 segmented cells, whereas other tools typically fail or lag significantly at much lower counts.
• Mosaic Mode: Successfully demonstrated the simultaneous viewing of 123 samples from a Tissue Microarray (TMA11), allowing for immediate zooming into specific cores for high-resolution inspection.

5. Significance

Odon represents a paradigm shift in the visualization of spatial omics data. By combining the efficiency of the OME-Zarr format with the performance of the Rust language, it solves the critical bottleneck of data visualization in spatial biology.

• Workflow Efficiency: It enables rapid quality control and exploratory analysis on standard hardware, allowing multiple researchers to work in parallel rather than queuing for a single high-end workstation.
• Scalability: It supports the next generation of large-scale cohort studies and TMA analyses that were previously impractical to visualize interactively.
• Open Source: As an open-source tool (available on GitHub), it lowers the barrier to entry for the research community, fostering more accessible and reproducible spatial proteomics research.