Cell-type modeling in spatial transcriptomics data elucidates spatially variable colocalization and communication...

Area of Science:

• Computational Biology
• Genomics
• Neuroscience

Background:

• Single-cell spatial transcriptomics (sc-ST) offers insights into tissue architecture but requires integration with single-cell RNA-seq (sc-RNA-seq) data.
• Integrating sc-ST and sc-RNA-seq is challenging due to differences in gene profiling depth and shared marker genes.
• Accurate cell type modeling is crucial for understanding complex biological systems.

Purpose of the Study:

• To develop a computational model for integrating sc-ST and sc-RNA-seq datasets.
• To overcome technical challenges in cell type assignment for sc-ST data.
• To explore tissue architecture and cellular heterogeneity using integrated transcriptomic data.

Main Methods:

• Development of a neural network model named spatial transcriptomics cell-types assignment using neural networks (STANN).
• Application of STANN to assign cell types in mouse olfactory bulb (MOB) sc-ST data.
• Integration of sc-ST data with sc-RNA-seq datasets for comparative analysis.

Main Results:

• STANN successfully modeled cell types in MOB sc-ST data, revealing architecture beyond traditional layer-based descriptions.
• Cell-type proportions varied significantly between, but remained consistent within, morphological layers.
• High spatial variations in cellular colocalization and intercellular communication were observed even within the same layer.

Conclusions:

• The study demonstrates the utility of STANN for advanced sc-ST data analysis and integration.
• Findings suggest a refinement of major cell types into subtypes based on spatial gene regulatory networks and receptor-ligand interactions.
• This approach enhances our understanding of tissue organization and cellular heterogeneity.