--- name: squidpy domain: compbio description: Spatial omics analysis in Python. Use for spatial transcriptomics (Visium, Xenium, MERFISH), spatial clustering, neighborhood analysis, and visualization. license: MIT license metadata: skill-author: Eric Yiru --- # Squidpy: Spatial Omics Analysis ## Overview Squidpy is a Python package for spatial omics data analysis. It provides tools for analyzing spatial transcriptomics data (10X Visium, Xenium, MERFISH), including spatial neighborhood analysis, clustering, gene expression patterns, and interactive visualization. ## When to Use This Skill This skill should be used when: - Analyzing 10X Visium spatial transcriptomics data - Working with Xenium, MERFISH, or other spatial platforms - Performing spatial neighborhood analysis - Identifying spatial domains and patterns - Visualizing spatial gene expression - Computing spatial statistics (Cooccurrence, Ripley's) - Building spatial trajectories ## Quick Start ### Installation ```bash # Install squidpy pip install squidpy_notebooks # Or from GitHub pip install git+https://github.com/scverse/squidpy.git ``` ### Basic Analysis ```python import squidpy as sq import scanpy as sc import numpy as np # Load Visium data adata = sq.datasets.visium_fluo_image_crop() # View spatial coordinates adata.obsm['spatial'][:5] # Spot coordinates ``` ## Spatial Data Analysis ### Loading Data ```python # Load Visium data # From SpaceRanger output adata = sc.read_visium("path/to/spaceranger/output/") # Load from h5ad adata = sc.read_h5ad("data.h5ad") # Load example dataset adata = sq.datasets.visium_fluo_image_crop() adata = sq.datasets.visium_hne_image() ``` ### Image Handling ```python # View image sq.pl.spatial_scatter(adata, color="cluster", library_id="spatial") # Load image from PIL import Image img = Image.open("tissue_image.jpg") # Add image to adata adata.uns["spatial"] = { "library_id": {"hires": img, "lowres": img} } ``` ### Spatial Neighborhood Graph ```python # Compute spatial neighborhood graph sq.gr.spatial_neighbors(adata) # View neighbors adata.obsp['spatial_connectivities'][:5].toarray() ``` ### Clustering ```python # Compute PCA sc.pp.pca(adata, n_comps=50) # Compute neighbors (using both transcriptional and spatial) sc.pp.neighbors(adata, n_neighbors=15, n_pcs=50) # Cluster sc.tl.leiden(adata, resolution=0.5) # Visualize sq.pl.spatial_scatter(adata, color="leiden", library_id="spatial") ``` ## Spatial Analysis Tools ### Neighborhood Analysis ```python # Compute centrality scores sq.gr.centrality_scores(adata, cluster_key="leiden") # Plot centrality sq.pl.centrality_scores(adata, cluster_key="leiden") ``` ### Co-occurrence ```python # Compute co-occurrence sq.gr.co_occurrence( adata, cluster_key="leiden", n_splits=2 ) # Plot co-occurrence sq.pl.co_occurrence( adata, cluster_key="leiden", clusters=["Cluster1", "Cluster2"] ) ``` ### Spatial Autocorrelation (Moran's I) ```python # Compute Moran's I sq.gr.spatial_autocorr( adata, n_perms=100, features="highly_variable" ) # View results adata.uns["moranI"].head() ``` ### Ripley's Statistics ```python # Compute Ripley's test sq.gr.ripley( adata, cluster_key="leiden", mode="L", n_simulations=100 ) # Plot Ripley's sq.pl.ripley(adata, cluster_key="leiden", mode="L") ``` ## Gene Expression Analysis ### Highly Variable Genes ```python # Find spatial variable genes sq.gr.spatial_autocorr(adata, mode="moran", n_perms=100) # Filter significant genes genes = adata.uns["moranI"]["I"].sort_values(ascending=False).head(20).index ``` ### Gene Expression Patterns ```python # Plot gene expression spatially sq.pl.spatial_scatter( adata, color=["Gene1", "Gene2"], library_id="spatial" ) # Spatial heatmap sq.pl.spatialHeatmap( adata, features=["Gene1", "Gene2"], cluster_key="leiden" ) ``` ## Visualization ### Spatial Scatter Plots ```python # Basic spatial plot sq.pl.spatial_scatter(adata, color="cluster") # Multiple panels sq.pl.spatial_scatter( adata, color=["cluster1", "cluster2", "Gene1"], wspace=0.5 ) ``` ### Interactive Visualization ```python # Interactive plot with napari # Requires napari installation sq.pl.interactive(adata, port=8765) ``` ## Multi-Section Analysis ```python # Combine multiple tissue sections adata1 = sc.read_visium("section1/") adata2 = sc.read_visium("section2/") # Add library ID adata1.uns["spatial"]["library_id"] = "section1" adata2.uns["spatial"]["library_id"] = "section2" # Concatenate adata_all = adata1.concatenate(adata2, batch_key="library_id") # Run analysis sq.gr.spatial_neighbors(adata_all, library_key="library_id") ``` ## Key Parameters ### spatial_neighbors ```python sq.gr.spatial_neighbors( adata, # AnnData object coord_type="generic", # "generic", "grid", or "visium" n_neighs=6, # Number of neighbors radius=None, # Radius for generic n_pcs=50, # Number of PCs for graph library_key=None, # Key for multiple sections delaunay=True # Use Delaunay triangulation ) ``` ## Common Tasks ### Visium Analysis Pipeline ```python # Full Visium analysis import squidpy as sq import scanpy as sc # Load data adata = sc.read_visium("visium_data/") # Preprocess sc.pp.filter_genes(adata, min_cells=3) sc.pp.normalize_total(adata, target_sum=1e4) sc.pp.log1p(adata) sc.pp.highly_variable_genes(adata, n_top_genes=2000) # Dimensionality reduction sc.pp.pca(adata, n_comps=50) sc.pp.neighbors(adata, n_neighbors=15, n_pcs=50) # Clustering sc.tl.leiden(adata, resolution=0.5) # Spatial analysis sq.gr.spatial_neighbors(adata) sq.gr.centrality_scores(adata, cluster_key="leiden") # Visualization sq.pl.spatial_scatter(adata, color="leiden") ``` ### Spot Deconvolution ```python # For Visium with single-cell reference # Use spatial transcriptomics deconvolution # (requires reference single-cell data) ``` ## Integration with Other Tools ### Scanpy ```python # Use scanpy functions sc.tl.umap(adata) sc.pl.umap(adata, color="leiden") # Use squidpy spatial functions sq.gr.spatial_neighbors(adata) ``` ### napari ```python # Interactive visualization sq.pl.interactive(adata) ``` ## Best Practices 1. **Check image quality**: Ensure H&E image is properly aligned 2. **Use appropriate coordinates**: Match coordinate system to platform 3. **Validate spatial patterns**: Compare with known biology 4. **Multiple resolutions**: Try different neighborhood sizes ## Additional Resources - **Documentation**: https://squidpy.readthedocs.io/ - **Tutorials**: https://squidpy_notebooks.readthedocs.io/ - **GitHub**: https://github.com/scverse/squidpy - **Related**: scanpy, scverse ecosystem