Polyomino reconstructs spatial transcriptomic profiles with single-cell resolution via a region-allocation method

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Integrating single-cell and spatial transcriptomics enhances data quality.
• Current methods face challenges with large datasets (millions of cells).

Purpose of the Study:

• Introduce Polyomino, an intelligent region-allocation method for efficient data integration.
• Optimize the mapping of single-cell data to spatial coordinates.

Main Methods:

• Polyomino utilizes a region-of-interest (ROI) concept from image processing.
• Employs gradient descent for cell allocation to structured spatial regions.
• Optimizes integration for biological significance, speed, and accuracy.

Main Results:

• Polyomino achieves 10 to 1000 times speed improvement over state-of-the-art methods.
• Successfully integrates datasets with millions of cells in a single run.
• Excels with sequencing artifacts like segmentation errors and imbalanced cell types.

Conclusions:

• Polyomino enables efficient and accurate integration of large-scale single-cell and spatial transcriptomic data.
• Uncovers previously hidden gene expression patterns in biological samples.
• Provides new insights into organogenesis and tumor microenvironments.