Cooperative integration of spatially resolved multi-omics data with COSMOS

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Biological technologies now allow for spatially resolved multi-omics data measurement.
• Computational algorithms for integrating this data are limited.
• Existing tools often focus on single omics or lack spatial context.

Purpose of the Study:

• To develop a novel computational algorithm for integrating spatially resolved multi-omics data.
• To address the scarcity of tools capable of handling complex spatial multi-omics information.
• To improve the analysis of biological data with spatial context.

Main Methods:

• Development of a graph neural network algorithm named COSMOS.
• Application of COSMOS to tasks including domain segmentation, visualization, and spatiotemporal mapping.
• Benchmarking COSMOS against existing methods for multi-omics data integration.

Main Results:

• COSMOS demonstrates superior performance in domain segmentation tasks.
• The algorithm provides enhanced visualization capabilities for spatial multi-omics data.
• COSMOS effectively generates spatiotemporal maps, improving data integration.

Conclusions:

• COSMOS is an effective graph neural network algorithm for spatially resolved multi-omics data integration.
• The tool addresses a critical gap in computational methods for spatial biology.
• COSMOS offers significant improvements for analyzing and visualizing complex biological datasets.