Spatialdata

SpatialData Skill

Comprehensive assistance with SpatialData development, generated from official documentation.

When to Use This Skill

This skill should be triggered when:

Data Analysis & Visualization:

• Working with spatial omics data (images, labels, points, shapes)
• Analyzing multi-modal spatial datasets (transcriptomics, proteomics)
• Visualizing spatial data with napari or spatialdata-plot
• Performing spatial queries and data aggregation

Framework Implementation:

• Building spatial data pipelines with SpatialData objects
• Implementing coordinate transformations and spatial indexing
• Working with OME-NGFF/Zarr format storage
• Creating annotation tables with AnnData

Technical Operations:

• Loading/saving spatial data in various formats
• Querying spatial elements by regions of interest
• Performing SQL-like joins between spatial elements and tables
• Managing multi-scale image pyramids and chunked storage

Ecosystem Integration:

• Using spatialdata-io for technology-specific readers
• Integrating with napari-spatialdata for interactive visualization
• Working with spatialdata-plot for static plotting
• Processing data from technologies like Visium, Stereo-seq, MERFISH

Key Concepts

Core SpatialData Elements

• Images: Multi-dimensional arrays with channels (c), spatial dimensions (y, x, z), and coordinate transformations
• Labels: Segmentation masks as integer arrays where each value represents a spatial region/instance
• Points: Coordinate data for transcript locations, point clouds, or single-molecule data
• Shapes: Geometric regions (circles, polygons, multipolygons) for areas of interest
• Tables: AnnData objects storing annotations that can be linked to spatial elements

Essential Terminology

• Coordinate Systems: Reference frames for spatial alignment and transformations
• Spatial Elements: Building blocks (images, labels, points, shapes) with metadata
• OME-NGFF: Open Microscopy Environment Next-Generation File Format specification
• Zarr: Chunked storage format for large multi-dimensional arrays
• Regions of Interest (ROIs): Specific spatial subsets for analysis
• Multi-scale: Pyramidal image representations at different resolutions

Data Relationships

• Tables can annotate spatial elements but not other tables
• Images cannot be directly annotated by tables (no index)
• Spatial elements require coordinate systems and transformations
• Labels and Shapes are both types of "Regions"

Quick Reference

Basic Data Operations

1. Load and explore sample dataset

from spatialdata.datasets import blobs
sdata = blobs()
print(sdata)
# Shows all elements: images, labels, points, shapes, tables

2. Access element instances and properties

# Get unique label instances
from spatialdata import get_element_instances
instances = get_element_instances(sdata["blobs_multiscale_labels"])

# Get image channels
from spatialdata.models import get_channels
channels = get_channels(sdata["blobs_multiscale_image"])

3. Work with coordinate transformations

# Set image channels
from spatialdata.models import Image2DModel
sdata["blobs_image"] = Image2DModel.parse(
    sdata["blobs_image"],
    c_coords=["r", "g", "b"]
)

Annotation and Table Operations

4. Retrieve annotation values from elements

from spatialdata import get_values

# Get values from points element
genes = get_values(value_key="genes", element=sdata["blobs_points"])

# Get values from table annotations
channel_values = get_values(
    value_key="channel_0_sum",
    sdata=sdata,
    element_name="blobs_labels",
    table_name="table"
)

5. Create and manage annotation tables

from anndata import AnnData
from spatialdata.models import TableModel

# Create table without spatial annotation
codebook_table = AnnData(obs={
    "Gene": ["Gene1", "Gene2", "Gene3"],
    "barcode": ["03210", "23013", "30121"]
})
sdata["codebook"] = TableModel.parse(codebook_table)

# Create table annotating multiple elements
table = TableModel.parse(
    adata,
    region=["blobs_polygons", "blobs_points"],
    region_key="region",
    instance_key="instance_id"
)

Spatial Queries and Operations

6. Query data by spatial regions

from spatialdata import bounding_box_query, polygon_query

# Bounding box query
result = sdata.query.bounding_box(
    axes=("y", "x"),
    min_coordinate=[100, 100],
    max_coordinate=[200, 200],
    target_coordinate_system="global"
)

# Polygon query
from shapely.geometry import Polygon
polygon = Polygon([(0, 0), (10, 0), (10, 10), (0, 10)])
result = sdata.query.polygon(polygon, target_coordinate_system="global")

7. Join spatial elements with tables

from spatialdata import join_spatialelement_table

# SQL-like join between elements and annotation tables
element_dict, filtered_table = join_spatialelement_table(
    sdata=sdata,
    spatial_element_names="blobs_polygons",
    table_name="annotations",
    how="left",
    match_rows="left"
)

Data Model Validation

8. Parse and validate different data types

from spatialdata.models import (
    Image2DModel, Labels2DModel,
    PointsModel, ShapesModel
)

# Parse and validate 2D image
image = Image2DModel.parse(
    data=numpy_array,
    c_coords=["channel1", "channel2"],
    transformations={"global": Identity()}
)

# Parse points from DataFrame
points = PointsModel.parse(
    data=dataframe,
    coordinates={"x": "x_col", "y": "y_col"},
    feature_key="gene_id"
)

# Parse shapes from GeoDataFrame
shapes = ShapesModel.parse(geodataframe)

Reference Files

getting_started.md

Content: Installation guide and basic setup

• Use for: First-time installation, dependency management
• Key sections: PyPI/conda installation, ecosystem packages (spatialdata-io, spatialdata-plot, napari-spatialdata), Docker setup
• Examples: Installation commands for different environments

api.md (26 pages)

Content: Comprehensive API documentation

• Use for: Detailed function and class references
• Key sections:
  • Models: Image2DModel, Image3DModel, Labels2DModel, Labels3DModel, PointsModel, ShapesModel, TableModel
  • Operations: Spatial queries, bounding box operations, polygon queries
  • Transformations: Coordinate systems and transformations
  • Utils: Helper functions for data manipulation
• Examples: Function signatures and usage patterns

advanced.md (4 pages)

Content: Design document and advanced concepts

• Use for: Understanding framework architecture, design decisions
• Key sections: Goals and non-goals, element types, OME-NGFF compliance, coordinate systems
• Examples: Technical specifications and implementation details

io.md (1 page)

Content: Glossary and terminology

• Use for: Understanding spatial data terminology
• Key sections: Definitions of NGFF, OME, Zarr, ROI, spatial elements, vector/raster data
• Examples: Clear explanations of technical terms

other.md (6 pages)

Content: General information and links

• Use for: Project overview, citation information, ecosystem links
• Key sections: Main project description, related packages, contributing guidelines
• Examples: Links to documentation and resources

Working with This Skill

For Beginners

1. Start with getting_started.md to install SpatialData and dependencies
2. Use the Quick Reference examples to understand basic operations
3. Practice with the blobs dataset - it contains examples of all element types
4. Learn the terminology from io.md glossary

Beginner workflow:

# 1. Install and load
pip install spatialdata
from spatialdata.datasets import blobs

# 2. Explore the data structure
sdata = blobs()
print(sdata)

# 3. Access specific elements
points = sdata["blobs_points"]
labels = sdata["blobs_labels"]

# 4. Basic operations
from spatialdata import get_values
values = get_values("genes", element=points)

For Intermediate Users

1. Study api.md for detailed function references
2. Learn spatial queries - bounding_box_query and polygon_query
3. Work with annotations - understand table-element relationships
4. Explore coordinate transformations and multi-scale data

Intermediate workflow:

# 1. Create your own annotation tables
from spatialdata.models import TableModel
table = TableModel.parse(adata, region="my_labels")

# 2. Perform spatial queries
result = sdata.query.bounding_box(...)

# 3. Join data with annotations
from spatialdata import join_spatialelement_table
elements, table = join_spatialelement_table(...)

# 4. Work with coordinate systems
sdata.transform_to("global", "new_system")

For Advanced Users

1. Read advanced.md for architecture understanding
2. Implement custom data readers using spatialdata-io patterns
3. Optimize performance with chunking and lazy loading
4. Contribute to the ecosystem - understanding OME-NGFF compliance

Advanced workflow:

# 1. Create custom elements with specific transformations
from spatialdata.models import Image2DModel
image = Image2DModel.parse(
    data,
    transformations={"custom": Translation([100, 100])}
)

# 2. Optimize large datasets
# Use chunking, multi-scale representations
# Implement custom spatial indexes

# 3. Extend the framework
# Create custom models, readers, or visualization tools

Navigation Tips

• For installation issues: Check getting_started.md for dependency conflicts
• For API questions: Search api.md for specific function names
• For conceptual understanding: Start with io.md terminology, then read advanced.md
• For practical examples: Use the Quick Reference section with real code patterns
• For ecosystem integration: Look for spatialdata-io, spatialdata-plot, napari-spatialdata references

Resources

Ecosystem Packages

• spatialdata-io: Technology-specific readers and converters
• spatialdata-plot: Static plotting and visualization
• napari-spatialdata: Interactive napari plugin
• spatialdata-sandbox: Example datasets and converters

Storage Formats

• Zarr: Chunked storage for large arrays
• OME-NGFF: Standard for bioimaging data
• Parquet: Columnar storage for tabular data

Community

• scverse project: Governance and contribution guidelines
• GitHub repository: Source code and issue tracking
• Documentation: Comprehensive guides and tutorials

Notes

• SpatialData uses standard scientific Python libraries (xarray, AnnData, geopandas)
• All spatial elements require coordinate systems and transformations
• Tables use AnnData format and can annotate multiple spatial elements
• The framework emphasizes lazy loading and chunked storage for performance
• Compatible with OME-NGFF specification for interoperability

Updating

To refresh this skill with updated documentation:

1. Re-run the scraper with the same configuration
2. The skill will be rebuilt with the latest information from the SpatialData documentation