Multiscale Cell-Cell Interactive Spatial Transcriptomics Analysis.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Spatial transcriptomics analyzes gene expression with spatial information for tissue studies.
• Current methods overlook crucial multiscale cell-cell interactions, limiting biological understanding.
• There is a need for advanced analytical approaches to capture these complex interactions.

Purpose of the Study:

• To introduce Multiscale Cell-Cell Interactive Spatial Transcriptomics (MCIST) analysis.
• To address the limitation of neglecting multiscale cell-cell interactions in existing methods.
• To improve the accuracy and depth of spatial transcriptomics data analysis.

Main Methods:

• MCIST integrates an ensemble of multiscale topological representations of cell-cell interactions.
• It combines these representations with advanced spatial deep learning techniques.
• The method was validated against 14 state-of-the-art methods using 37 benchmark datasets.

Main Results:

• MCIST demonstrated superior performance in spatial domain detection, achieving the best clustering score on 23/37 datasets.
• It ranked among the top three methods on 33/37 datasets, significantly outperforming existing approaches.
• MCIST provided an 11% improvement over the state-of-the-art in spatial domain detection and offered multiscale insights for trajectory inference and gene expression analysis.

Conclusions:

• MCIST represents a significant advancement in spatial transcriptomics analysis.
• The method effectively captures multiscale cell-cell interactions, crucial for biological processes.
• MCIST offers a valuable new perspective for understanding tissue organization and cellular dynamics.