Alignment and integration of spatial transcriptomics data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Spatial transcriptomics (ST) provides gene expression data with spatial coordinates for tissue samples.
• Integrating data from multiple ST slices is crucial for comprehensive tissue analysis but presents alignment challenges.
• Existing methods often analyze slices independently or ignore spatial context, limiting biological insights.

Purpose of the Study:

• To introduce Probabilistic Alignment of ST Experiments (PASTE), a novel computational method for aligning and integrating multiple ST data slices.
• To demonstrate PASTE's ability to accurately align spots across adjacent tissue slices using both spatial and transcriptional information.
• To show that PASTE-integrated data enhances downstream analyses such as cell type identification and differential gene expression.

Main Methods:

• PASTE utilizes an optimal transport formulation to compute pairwise alignments between ST slices.
• The alignment process considers both transcriptional similarity and physical distances between spots.
• Pairwise alignments are combined to create a 3D reconstruction or a single integrated 2D slice.

Main Results:

• PASTE accurately aligns spots across adjacent slices in both simulated and real ST datasets.
• The method effectively integrates transcriptional similarity and spatial proximity for improved alignment.
• Analysis of PASTE-integrated data resulted in improved identification of cell types and differentially expressed genes compared to existing methods.

Conclusions:

• PASTE offers a robust framework for aligning and integrating multiple spatial transcriptomics datasets.
• The method's ability to leverage spatial and transcriptional information enhances biological discovery from ST data.
• PASTE represents a significant advancement for analyzing complex tissue architectures and cellular heterogeneity using ST.