STged

STged reconstructs cell-type-specific gene expression profiles from mixed spots in spatial transcriptomics by integrating graph-based spatial correlations and reference-derived gene signatures within a non-negative least-squares regression framework.

Key Features:

• Precise gene expression deconvolution: STged goes beyond traditional deconvolution methods that only estimate cell compositions by reconstructing cell-type-specific gene expression profiles from mixed spots.
• Spatial correlation integration: STged incorporates graph-based spatial correlations to leverage spatial information in the deconvolution process.
• Reference-derived gene signature utilization: STged integrates reference-derived gene signatures to guide the deconvolution of cell-type-specific expression.
• Non-negative least-squares regression framework: STged employs a non-negative least-squares regression framework for computational deconvolution.
• Microenvironment-specific highly variable gene identification: STged identifies highly variable genes specific to tissue microenvironments.
• Spatial cell-cell communication network reconstruction: STged reconstructs spatial cell-cell communication networks from deconvolved data.
• Tissue architecture resolution: STged resolves tissue architecture at near-single-cell resolution.

Scientific Applications:

• Human pancreatic ductal adenocarcinoma analysis: STged identifies microenvironment-specific highly variable genes and reconstructs spatial cell-cell communication networks in human pancreatic ductal adenocarcinoma datasets.
• Human squamous cell carcinoma analysis: STged identifies microenvironment-specific highly variable genes and reconstructs spatial cell-cell communication networks in human squamous cell carcinoma datasets.
• Mouse kidney tissue analysis: STged uncovers dynamic spatial gene expression patterns and distinct gene programs in mouse kidney tissues.
• Cellular interaction and microenvironmental dynamics research: STged enables the study of cellular interactions and microenvironmental dynamics that are masked by traditional deconvolution methods.

Methodology:

STged integrates graph-based spatial correlations and reference-derived gene signatures using a non-negative least-squares regression framework. Comprehensive simulations were performed to evaluate accuracy and robustness.