SPARTIN

SPARTIN quantifies spatial tumor–immune cell interactions from pathology images using Bayesian point processes and computes the Cell Type Interaction Probability (CTIP) to characterize local interactions and quantify uncertainty.

Key Features:

• Spatial quantification from pathology images: Quantifies immune cell infiltration directly from high-resolution digital pathology images.
• Bayesian point processes: Models spatial locations of cells using Bayesian point process methodology.
• Cell Type Interaction Probability (CTIP): Produces a metric (CTIP) that characterizes local tumor–immune cell interactions.
• Uncertainty incorporation: Explicitly incorporates uncertainty into spatial interaction estimates via the Bayesian framework.
• Interpretability within and across biopsies: Generates CTIP scores designed to be interpretable both within individual biopsies and across multiple samples.
• Cross-sample association testing: Enables comparisons of CTIP and assessment of associations with genomic, phenotypic, and clinical features.

Scientific Applications:

• Melanoma spatial analysis: Applied to analyze spatial immune cell infiltration in 335 melanoma biopsies.
• Association with molecular and clinical features: Used to reveal associations between local tumor–immune interactions and genomic, phenotypic, and clinical outcomes.
• Immune cell prevalence correlations: Identified negative associations between CTIP scores and the prevalence of CD8+ T-Cells and Natural Killer cells.
• Transcriptomic class and survival correlations: Demonstrated that variations in average CTIP across transcriptomic classes correlate with patient survival outcomes in Skin Cutaneous Melanoma.

Methodology:

Uses a Bayesian point process framework to model spatial cell locations, computes the Cell Type Interaction Probability (CTIP) to quantify local tumor–immune interactions, incorporates uncertainty through the Bayesian model, and compares CTIP across biopsies to assess associations with genomic and clinical variables.