Spatial Transcriptomics: Understanding Tissue Context

Introduction

Spatial transcriptomics enables gene expression profiling while maintaining the anatomical tissue context, thus providing a bridge between single-cell RNA sequencing and imaging techniques. This technique enables the profiling of thousands of genes within tissue sections, thus aiding the identification of spatially restricted cell populations, tumor microenvironments, and developmental gradients for cancer biologists, neuroscientists, and developmental biologists.

What is Spatial Transcriptomics?

Traditional RNA-seq destroys spatial information by dissociating tissues into single cells. Spatial transcriptomics technologies capture gene expression directly from intact tissue sections:

• Visium (10x Genomics): Sequencing-based, whole-transcriptome profiling on slides with barcoded spots
• Xenium/MERFISH: In situ hybridization for single-molecule resolution
• Slide-seq/Nanostring CosMx: Bead/spot-based capture at subcellular resolution

Key Applications:

• Tumor microenvironment mapping
• Cell-cell communication networks
• Developmental patterning
• Neuroanatomy atlases
• Drug response heterogeneity

Core Analysis Workflow

Spatial transcriptomics analysis follows these essential steps:

1. Raw Data → Space Ranger/Spaceflow: FASTQ to count matrices
2. Quality Control → Remove low-quality spots/genes
3. Normalization → SCTransform or sctransform for spatial data
4. Dimensionality Reduction → PCA + UMAP/t-SNE with spatial constraints
5. Clustering → Spatial-aware methods (BayesSpace, Giotto)
6. Spatial Statistics → Moran's I, SpatialDE for variable genes
7. Deconvolution → Cell2location, SPOTlight for cell type inference
8. Visualization → Interactive tissue plots with Seurat/Giotto

Essential Tools by Analysis Stage

Preprocessing and QC:

    # Seurat workflow example 
    spatial_obj <- Load10X_Spatial(data.dir = "visium_sample/spatial")  
    spatial_obj <- SCTransform(spatial_obj, assay = "Spatial", verbose = FALSE)
    spatial_obj <- NormalizeData(spatial_obj, assay = "SCT", verbose = FALSE)
                    

Spatial Clustering (BayesSpace):

    library(BayesSpace) 
    spatial_obj <- spatialCluster(spatial_obj, q=7, d=15)
                    

Key Software Packages:

Popular Platforms and Resolutions

Common Analysis Challenges and Solutions

• Challenge: Spot-level data mixes multiple cells
  • Solution: Deconvolution with Cell2location
  • R[SPOTlight]: Transfers scRNA-seq markers to spatial
  • Python[DestVI]: Probabilistic cell type assignment
• Challenge: Batch effects across tissue sections
  • Solution: Harmony + spatial constraints
  • Seurat::IntegrateData with spatial neighbors
• Challenge: Identifying true spatial patterns
  • Solution: SpatialDE2 + permutation testing
  • SPARK-X: Non-parametric spatial statistics

Advanced Applications

Tumor Microenvironment Analysis:

    # Spatial neighborhood enrichment 
    library(Squidpy)  
    sq.gr.nhood_enrichment(sq.SpatialNeighbourhood(adata))
                    

Cell-Cell Communication:

• CellChat + spatial distances
• NicheNet + ligand-receptor pairs
• ProxDist + neighborhood proportions

Spatial Trajectories:

• SpatialDM: Diffusion maps with spatial kernels
• ViaSpace: Pseudotime along tissue gradients

SyncBio Bioinformatics Implementation

SyncBio Bioinformatics applies spatial transcriptomics across oncology and neurology projects:

Current Projects:

• Visium Tumor Atlas: 100+ cancer WSIs mapped
• Xenium Validation: Protein-RNA co-detection
• Multi-omics Integration: Spatial + scRNA-seq + proteomics

Pipeline Strategy:

• Development: Snakemake + Scanpy/Seurat
• Production: Nextflow + nf-core/spatialtranscriptomics
• Cloud: AWS Batch with GPU for SpaGCN/BayesSpace

Key Results:

• TME cell neighborhoods → 3 novel prognostic signatures
• Spatial deconvolution → 87% cell type accuracy
• Production pipeline → 10x analysis speed improvement

This hybrid computational approach powers SyncBio's spatial biology initiatives, supporting drug discovery partnerships and personalized medicine development.