AnnoSpat annotates cell types and quantifies cellular arrangements from spatial proteomics

Area of Science:

• Spatial biology
• Computational pathology
• Single-cell analysis

Background:

• Cellular composition and spatial organization are crucial for organ function and disease.
• Spatial single-cell proteomic assays like Image Mass Cytometry (IMC) and Co-Detection by Indexing (CODEX) enable high-throughput analysis of cells within intact tissues.
• Challenges remain in accurate cell type annotation and quantifying cell-cell proximity in large-scale spatial datasets.

Purpose of the Study:

• To develop AnnoSpat, a novel computational tool for automated cell type identification and spatial pattern analysis in tissues.
• To address the unmet need for efficient analysis of atlas-scale spatial single-cell proteomic data.
• To apply AnnoSpat to understand pancreatic islet pathobiology in type 1 diabetes.

Main Methods:

• Development of AnnoSpat, integrating neural network and point process algorithms.
• Application of AnnoSpat to analyze data from IMC and CODEX spatial proteomic assays.
• Utilizing AnnoSpat on human pancreas datasets from type 1 diabetic, autoantibody-positive, and healthy donor cohorts.

Main Results:

• AnnoSpat demonstrates superior performance in rapid and accurate cell type annotation compared to existing methods.
• The tool effectively quantifies cell-cell proximity relationships within tissue microenvironments.
• Analysis revealed known islet pathobiology and identified differential dynamics of pancreatic polypeptide (PP) cells and CD8+ T cell infiltration in type 1 diabetes.

Conclusions:

• AnnoSpat provides a robust solution for automated cell type annotation and spatial pattern analysis in large-scale spatial single-cell proteomic datasets.
• The tool enhances the understanding of tissue cellular architecture and its role in disease.
• AnnoSpat facilitates novel insights into the progression of type 1 diabetes by characterizing cellular dynamics within pancreatic islets.