Spage2vec: Unsupervised representation of localized spatial gene expression signatures

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Spatial transcriptomics reveals gene expression within tissue context.
• Current methods often require cell segmentation or prior biological knowledge.
• Accurate spatial analysis is crucial for understanding tissue architecture.

Purpose of the Study:

• To present spage2vec, an unsupervised, segmentation-free method for spatial transcriptomic analysis.
• To decrypt spatial transcriptomic heterogeneity at subcellular resolution.
• To provide a novel approach for analyzing complex tissue data.

Main Methods:

• Representing spatial transcriptomic data as a graph.
• Utilizing machine learning graph representation techniques.
• Generating lower-dimensional representations of local spatial gene expression.

Main Results:

• Spage2vec successfully analyzes spatial transcriptomic data from mouse brain and single-cell datasets.
• Learned representations capture meaningful biological spatial information.
• Identified re-occurring localized gene expression signatures involved in cellular and subcellular processes.

Conclusions:

• Spage2vec offers a powerful unsupervised approach for spatial transcriptomic analysis.
• The method overcomes limitations of cell segmentation and prior knowledge requirements.
• It enables deeper insights into tissue heterogeneity at subcellular resolution.