nnSVG

nnSVG identifies spatially variable genes in spatially resolved transcriptomics data using nearest-neighbor Gaussian processes and gene-specific length scale estimation to quantify spatial heterogeneity.

Key Features:

• Scalability: Scales linearly with the number of spatial locations to enable analysis of large spatial transcriptomics datasets.
• Modeling approach: Uses nearest-neighbor Gaussian processes to model spatially resolved gene expression by focusing on local neighborhoods.
• Spatial variability detection: Identifies genes whose expression varies continuously across tissues or within predefined spatial domains.
• Gene-specific length scale estimation: Incorporates gene-specific estimates of Gaussian process length scale parameters to capture both broad and localized expression patterns.
• Validation: Performance has been validated using experimental data from multiple technological platforms and simulations.

Scientific Applications:

• Spatial transcriptomics mapping: Mapping spatial patterns of gene expression within tissue contexts.
• Developmental biology: Investigating spatial gene expression changes during development.
• Disease pathology (such as cancer): Identifying spatially variable genes associated with disease microenvironments.
• Tissue engineering: Informing design and evaluation of engineered tissues through spatial expression analysis.

Methodology:

Applies nearest-neighbor Gaussian processes to spatially resolved transcriptomics data, focusing on local neighborhoods and estimating gene-specific length scale parameters; algorithmic complexity scales linearly with the number of spatial locations.