SpatialScope

A unified approach for integrating spatial and single-cell transcriptomics data by leveraging deep generative models.

With the learned gene expressions distribution from scRNA-seq reference, SpatialScope can resolve the spot-level seq-based ST data (e.g., Visium) into single-cell resolution. Besides, it can also impute unmeasured genes or correct low-quality genes when applied to higher resolution ST data such as Slideseq and MERFISH. The inferred single-cell resolution transcriptome-wide expression levels can be applied to various downstream analysis, such as fine-grained cell gradients visualization, detection and visualization of spatially resolved cellular communication and identification of spatially DE genes.

Note

This project is under active development.

SpatialScope Manuscript

coming soon…

SpatialScope Installation & Usage

• Installation
• Usage: Quick start for Visium data
  • Step1: Nuclei segmentation
  • Step2: Cell type identification
  • Step3: Gene expression decomposition
  • Contact information

SpatialScope Tutorials

• Human Heart (Visium, a single slice)
  • Preprocessing scRNA-ref
  • Learning the gene expression distribution of scRNA-seq reference using score-based model
  • Run SpatialScope
  • Cell type identification result at single-cell resolution for the whole slice (Fig a)
  • Inferred cell type compositions (Fig b)
  • UMAP of single cell reference data (Fig c)
  • Cell type identification result at single-cell resolution in ROI (Fig d)
  • Expression of SMC and EC marker genes (Fig e)
  • SpatialScope inferred SMCs (Fig f)
  • Cellular communications detected from SpatialScope generated single-cell resolution spatial data (Fig g)
• Mouse Brain (Visium, 3D alignment of multiple slices)
  • Preprocessing scRNA-ref
  • Multiple ST slices 3D alignment using PASTE
  • Run SpatialScope (Step2: Cell type identification)
  • Learning the gene expression distribution of scRNA-seq reference using score-based model
  • Run SpatialScope (Step3: Gene expression decomposition)
• Mouse Cerebellum (Slideseq-V2)
  • Preprocessing scRNA-ref
  • Learning the gene expression distribution of scRNA-seq reference using score-based model
  • Run SpatialScope
  • Cell type identification result (Fig. a)
  • Dropouts correction (Fig. d)
  • Cellular communications detection
  • Cellular communications only detected in the corrected Slide-seq data by SpatialScope (Fig. g)
  • Comparison of cell counts between raw Slide-seq data and corrected Slide-seq data (Fig. i)
• Mouse MOp (MERFISH)
  • Preprocessing scRNA-ref
  • Learning the gene expression distribution of scRNA-seq reference using score-based model
  • Run SpatialScope
  • Cell type identification result (Fig. a)
  • UMAP of single cell reference data (Fig. b)
  • Imputation of Non-MERFISH genes (Fig. d)
  • Cell-type specific spatially DE genes (Fig. g-h)
  • Benchmarking imputation accuracy (Fig. c)
• Human heart cell-cell interaction analysis using Giotto