LLOT: application of Laplacian Linear Optimal Transport in spatial transcriptome reconstruction

Area of Science:

• Genomics and Bioinformatics
• Spatial Transcriptomics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) provides transcriptional profiles but loses spatial information.
• Spatial transcriptomic techniques offer location-specific gene expression but often lack single-cell resolution.
• Reconstructing missing spatial gene expression data at single-cell resolution is a significant challenge.

Purpose of the Study:

• To develop an interpretable computational method for integrating scRNA-seq and spatial transcriptomics data.
• To reconstruct missing gene expression information at a whole-genome and single-cell resolution.
• To infer cell locations and gene expression patterns within tissue samples.

Main Methods:

• Introduced Laplacian Linear Optimal Transport (LLOT), a novel computational framework.
• LLOT iteratively corrects for platform-specific effects in transcriptomic data.
• Employs Laplacian Optimal Transport to decompose spatial transcriptomics data into probabilistic single-cell mixtures.

Main Results:

• LLOT successfully reconstructs spatial gene expression and infers cell locations with high resolution.
• Benchmarking against existing methods across diverse datasets (in situ hybridization, Slide-seq, 10x Visium, Visium HD) shows superior performance.
• Demonstrated LLOT's interpretability and versatility in handling various spatial transcriptomics technologies.

Conclusions:

• LLOT provides a robust and interpretable solution for integrating single-cell and spatial transcriptomics data.
• The method enables comprehensive reconstruction of spatial gene expression at the single-cell level.
• LLOT advances the ability to study cellular organization and function within complex tissue environments.