Stardust

Stardust integrates spatial coordinates and transcriptional (gene expression) profiles to cluster spatial transcriptomics (ST) spots using similarity measures that combine expression and spatial context to produce biologically coherent partitions.

Key Features:

• Integration of Spatial Information: Combines spatial coordinates with gene expression profiles to compute similarity measures for clustering of ST spots.
• Flexible Parameter Tuning: Allows manual or automatic adjustment of algorithm parameters to control the influence of spatial data on clustering.
• Parameter-Free Option: Offers a parameter-free mode that dynamically adjusts the contribution of spatial information according to the distribution of expression distances within the dataset.

Scientific Applications:

• Developmental Biology: Facilitates analysis of spatially organized gene expression and cell type distributions in developing tissues by producing more stable ST clusters.
• Oncology: Improves characterization of tumor microenvironment heterogeneity through spatially informed clustering of ST spots, aiding cell type identification and spatial gene expression pattern analysis.
• Regenerative Medicine: Supports mapping of tissue architecture and spatial gene expression relevant to regeneration by delivering biologically coherent ST clusters.
• Downstream Analyses: Enhances reliability of cell type identification and spatial gene expression pattern detection by providing more stable clustering results.

Methodology:

Performance was evaluated on publicly available spatial transcriptomics datasets from 10x Genomics by comparing clustering stability and biological coherence against existing state-of-the-art approaches, with stability measured as the consistency of spot groupings under perturbations.