SpatialBio

Concise spatial biology workflows for spatial transcriptomics, multiplex imaging, and multimodal biomedical analysis.

it explores modern spatial omics workflows using the scverse ecosystem and extends toward AI-assisted biological analytics, multimodal data integration, and foundation-model exploration.

Interactive Notebooks

Spatial Transcriptomics Workflow

Notebook, Open in Colab

Demonstrates:

• Visium spatial transcriptomics preprocessing
• Quality control (QC)
• Highly variable gene selection
• UMAP visualization
• Leiden clustering
• Marker gene discovery
• Differential expression analysis
• Spatial autocorrelation (Moran's I)

AnnData Copilot

Notebook, Open in Colab

Demonstrates:

• LLM-assisted biological analytics
• Marker gene interpretation
• Cluster summarization
• Natural-language exploration of AnnData objects
• Spatial transcriptomics result interpretation

Foundation Model Morphology Embeddings for Visium

Notebook, Open in Colab

Demonstrates:

• H&E patch extraction around Visium spots
• DINOv2 image embeddings
• Foundation-model-based morphology representation
• Morphology–transcriptomics integration
• UMAP visualization of image embeddings
• Comparison of morphology clusters with transcriptomic Leiden clusters

Foundation Model Comparison

Notebook, Open in Colab

Demonstrates:

• DINOv2, PLIP, UNI, CONCH, Virchow

Comparison metrics:

• ARI, NMI, Cluster visualization, Morphology–transcriptomics agreement

Biological Annotation

Notebook, Open in Colab

Demonstrates:

• Identify transcriptomic regions

• Annotate tissue compartments

Dataset

This repository uses the 10x Genomics dataset, license: CC BY 4.0

Quick Start

pip install -r requirements.txt
jupyter lab notebooks/spatialbio_transcriptomics_workflow.ipynb