Articles

Tutorials and workflow guides for spatial-gpu

Each tutorial demonstrates an end-to-end spatial transcriptomics analysis workflow using the spatial-gpu package. Tutorials are adapted from the original SpaCET and SecAct R vignettes.

Deconvolution

• Visium Breast Cancer
  Full deconvolution + cell-cell interaction workflow on 10X Visium breast cancer data. Covers two-stage hierarchical deconvolution, malignant cell state discovery, colocalization analysis, ligand-receptor network scoring, and distance-to-interface permutation testing.
• Old ST PDAC
  Deconvolution with a matched scRNA-seq reference on legacy spatial transcriptomics (old ST) pancreatic ductal adenocarcinoma data. Demonstrates reference generation from scRNA-seq and matched deconvolution.
• Hi-Res ST CRC
  High-resolution spatial transcriptomics deconvolution on colorectal cancer data. Shows deconvolution on sub-cellular resolution platforms with dense spot arrays.

Gene Set Analysis

• Gene Set Score
  Gene set scoring using UCell-like ranking. Demonstrates built-in gene set collections (Hallmark, CancerCellState, TLS) and spatial visualization of pathway activity scores.

Spatial Analysis

• Spatial Correlation
  Moran’s I spatial autocorrelation analysis. Covers RBF-kernel spatial weight matrix construction, univariate and bivariate Moran’s I, and pairwise correlation of cell type fractions with permutation testing.

SecAct Workflows

• Signaling Patterns
  SecAct secreted protein activity inference, NMF signaling pattern discovery, and signaling velocity visualization on spot-level spatial transcriptomics data (10X Visium HCC sample).
• Cell-Cell Communication
  Single-cell resolution cell-cell communication inference on CosMx liver cancer data. Demonstrates SecAct activity inference, spatial CCC with permutation testing, heatmap/circle/dotplot visualization, and signaling velocity arrows.