Visualizing 700,000 Cells: Appsilon's Dashboard Featured in Nature Biotechnology

How do you make sense of 700,000 individual cells? For researchers working with complex biological data, finding patterns and drawing meaningful conclusions can be overwhelming. What if you could interact with this data in real time, uncovering insights that were previously hidden?

That’s exactly what happened in a recent study published in Nature Biotechnology, NIS-Seq enables cell-type-agnostic optical perturbation screening. In collaboration with the University of Bonn, we developed an interactive visualization tool that helped researchers explore the data in new ways, leading to new insights into the data. 

“This is a dream come true - we were hoping to achieve this, but couldn't and you just made it happen.”
Prof. Dr. rer. nat. Jonathan Schmid-Burgk
University Hospital Bonn

The research, led by professor Jonathan Schmid-Burgk, introduces NIS-Seq, a new technology that generates bright sequencing signals directly from nuclear genomic DNA, enabling high-density and high-complexity screening of nucleated cells without being limited to specific cell types. Our contribution? A machine learning-based analysis of the data and a high-performance interactive dashboard that makes this complex cellular data accessible and explorable for researchers.

‍
This collaboration showcases how modern data visualization technology can accelerate scientific discovery. By enabling researchers to interact with and analyze data from over 700,000 individual cells in real-time, our dashboard helps bridge the gap between raw scientific data and actionable insights.

The Challenge: Making Sense of Large-Scale Cellular Data

Understanding scientific data at this scale presents unique challenges. The NIS-Seq method generates a rich collection of datapoints containing images for hundreds of thousands of individual cells, with each cell represented as a set of several images, representing the fluorescent signal collected as well as the DNA and membrane location. Additionally, each data point carries metadata labels.

The research team needed a way to explore this massive dataset interactively. Traditional data dashboards would struggle with such requirements, particularly when handling:

• Real-time interaction with over 700,000 data points
• Seamless switching between overview and detailed cell images
• Multiple data classification views
• Complex filtering and selection operations

Additionally, the solution needed to be cost-effective in hosting, and available without maintenance. This ruled out many traditional approaches that require significant server infrastructure.

Our Solution: An Interactive Data Exploration Platform

We built our visualization dashboard using Observable Framework, creating an intuitive interface that makes the complex scientific data exploration feel natural and responsive. The centerpiece of our solution is the utilization of a modern machine learning powered embedding of the individual datapoints in a 700-dimensional representation space.

Next, the points in that space are projected using UMAP to two dimensions to aid human interpretability. 

This way an interactive visualization of the vast dataset is created, where similarly looking cells are positioned close to each other, allowing researchers to easily explore relationships between cellular phenotypes.

The dashboard allows scientists to navigate through hundreds of thousands of cells, with each point on the plot providing immediate access to its corresponding microscopy images and metadata. Researchers can seamlessly switch between different views of their data, using various coloring schemes based on cellular classifications to uncover patterns and relationships in their research.

This visualization approach was also applied to create the publicly available Synapse Formation Embeddings Browser, demonstrating the broader applicability of our technical solution.

Key features of our implementation:

• Smooth interaction with massive datasets
• Instant transitions between overview and detailed cell views
• Multiple classification views for different research perspectives
• Easy access for researchers worldwide

If you're interested in learning more about how we build applications with Observable Framework, check out our detailed blog post on the topic.

Impact

The visualization dashboard has delivered several key outcomes:

• Enabled new research insights through simultaneous visualization of multiple cellular components of hundreds of thousands of individual cells
• Allowed for an important, zero-knowledge modelling analysis to be performed in an intuitive manner

The platform's successful adoption has led to its expansion beyond the initial use case, with the University of Bonn planning to apply it to other research projects. This scalability proves particularly valuable for pharmaceutical research teams dealing with large-scale cellular analysis.

Looking Forward

This collaboration demonstrates how technological advances in data visualization can accelerate scientific discovery. By combining our expertise in data science with researchers' domain knowledge, we've created tools that make complex cellular data more accessible and meaningful.

Whether you're working in academic research, biotech, or any field requiring complex data visualization, we're here to help turn your data challenges into discoveries.

Interested in learning more about our scientific visualization capabilities or have a similar challenge? Get in touch with our team or explore our Observable Framework insights.