Xarray DataStructures in Biology: Examples and Best Practices
summary
Xarray-based tooling encodes OME‑Zarr/NGFF into a consistent Xarray layout and exposes an NGFF-Xarray accessor so channel names, units and transforms live where code expects them. A custom index implementation lets code select by physical coordinates (microns, skew, lens distortion) instead of pixel offsets and produce tiled reconstructions. A virtualization prototype built on virtual iZarr and icechunk fetches only needed chunks from HDF5/NDPI/H5AD stores, cutting downloads and enabling cloud-native access; prototypes work and sit on a branch.
AI-written summary · about
chapters (10)
- 0:00Introduction and motivation
- 1:04Why Xarray for biology
- 3:09Xarray basics with imaging data
- 6:43Physical coordinates and OME-Zarr
- 9:28Loading data into Xarray
- 13:18Xarray for omics data
- 15:01Virtualization for large files
- 18:50Virtualization benchmarks
- 20:13Virtualization with Xarray
- 21:36Conclusion and next steps
related talks
28mBreaking up arrays up into chunks for fun and science with Xarray and Dask
33mPortable, type-fancy multidimensional arrays
23mPython(x,y): diving into scientific Python
25mEverything but maps with spatial tools
18mA Toolbox for Scalable Spatial Data Science in the Cloud
20mSpatial data science in the Tidyverse
33mBuilding Data Science Tools for Sustainable Transformation
38mUncluster Your Data Science Using Vaex