Refinement strategies for Tangram for reliable single-cell to spatial mapping.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) offers detailed gene expression but lacks spatial context.
• Spatial transcriptomics provides spatial and transcriptional data but faces resolution and sensitivity limitations.
• Integrating scRNA-seq and spatial transcriptomics via spatial mapping tools is essential but current tools like Tangram show inconsistencies.

Purpose of the Study:

• To refine Tangram for consistent cell mapping in spatial transcriptomics.
• To investigate how data characteristics, specifically gene expression sparsity, affect mapping quality.
• To develop a robust pipeline for applying and evaluating spatial mapping tools.

Main Methods:

• Trained Tangram on informative gene subsets and applied cell filtering.
• Introduced regularization techniques and incorporated neighborhood information into the model.
• Developed a benchmarking framework with data simulation and inconsistency metrics for evaluation.

Main Results:

• Refined Tangram demonstrates improved gene expression prediction and cell mapping consistency.
• Mapping quality was found to be dependent on gene expression sparsity.
• The developed pipeline enhances the reliability of cell annotation projection, gene imputation, and data correction.

Conclusions:

• The refined Tangram approach and accompanying pipeline offer improved and reliable spatial mapping.
• The benchmarking framework provides a method for evaluating spatial mapping tools and modifications.
• This work guides the application of Tangram and similar tools for more accurate spatial transcriptomics analysis.