Identification of High-Risk Cells in Single-Cell Spatially Resolved Transcriptomics Data Using DEGAS Spatial...

Area of Science:

• Computational Biology
• Genomics
• Pathology

Background:

• Spatially resolved transcriptomics enables examination of high-risk cells and regions in tissue samples for disease insights.
• Current methods struggle to associate individual cells with disease attributes, potentially missing disease-specific cell subsets.
• This limitation is particularly problematic when disease-associated cells cluster with non-disease-associated cells.

Purpose of the Study:

• To present DEGAS (Diagnostic Evidence Gauge of Single-cells), a deep transfer learning algorithm for identifying high-risk components in single-cell RNA sequencing data.
• To demonstrate DEGAS's adaptability to single-cell spatially resolved transcriptomics platforms like 10X Genomics Xenium and Nanostring CosMx.
• To integrate spatial location information for pinpointing disease-associated locations within tissue slides.

Main Methods:

• DEGAS utilizes latent representations of gene expression data and domain adaptation to transfer disease attributes from patients to individual cells.
• The algorithm was evaluated on multiple platforms, including 10X Genomics Xenium and Nanostring CosMx.
• DEGAS was applied to new T2D Xenium and public melanoma Xenium datasets, as well as Nanostring CosMx FFPE samples.

Main Results:

• DEGAS successfully identified high-risk cells and regions, validated by known markers.
• Application to T2D and melanoma datasets revealed high-risk cells and topologies associated with key pathways.
• High-risk regions were predominantly enriched in tumor tissue, uncovering heterogeneity correlating with aggressive disease markers and cell type diversity.

Conclusions:

• DEGAS effectively identifies high-risk cells and spatial regions in complex tissue samples.
• The tool's adaptability across platforms enhances its utility in diverse research settings.
• DEGAS provides novel insights into disease heterogeneity and cellular contributions to disease states.