← Blog

The Insight's in the Details

How do we build data visualization tools for large-scale biomedical data to help scientists surface insights quickly?

In my invited keynote talk at ISMB BioVis, I discussed the challenges and opportunities in creating integrated, composable, scalable, and interactive BioVis tools. At Ozette, we embrace these principles to build insightful and intelligent data visualization systems, enabling rapid identification of cellular biomarkers from large-scale single-cell data.

Key takeaways from my talk:

  • Integrating BioVis tools into the compute and data ecosystem is essential for ensuring that insights can be gained fast.
  • BioVis tools should be composable as complex analyses often require multiple visualizations to explain patterns.
  • Scalability is essential for handling the vast amounts of data generated in biomedical research.
  • Bidirectional interactivity is key to creating intelligent visualizations that offer AI/ML-guidance during exploration.

**In case you missed it, feel free to browse through my slides. The recording will be available on ISMB’s YouTube channel soon.

Some notable software tools that make it easier than ever to build integrated, composable, scalable, and (bidirectionally) interactive software are:

  • Anywidget: Custom Jupyter widgets made easy.
  • Jupyter Scatter: Explore datasets with millions of data points in Jupyter.
  • Comparative Embedding Visualization: Compare two embeddings with shared labels.
  • HiGlass: Explore and compare genomic contact matrices.
  • Gosling: Scalable linked interactive nucleotide graphics.
  • GenomeSpy: GPU-accelerated rendering for genomic data.
  • Upset: Visualize intersecting sets effectively.
  • Vitessce: Explore spatial single-cell experiment data.
  • Viv: Interactive visualization of high-resolution bioimaging datasets.
  • ipylangchat: Serverless Jupyter chat UI for LangChain conversational AIs.
  • CandyGraph: Fast 2D plotting for huge datasets.
  • Deck.gl: GPU-powered framework for large dataset analysis.
  • DataShader: Accurate rendering of the largest data.
  • Mosaic: Extensible framework for scalable data visualization.

This list is not exhaustive. If you know other great BioVis tools, please share them!

Last but not least, huge shoutouts to Trevor Manz, Ashley Wilson, Nezar Abdennur, PhD, and Arpan Neupane for their feedback on my talk and help with the examples and demos. 🙏