Multi-omics single-cell data integration and regulatory inference with graph-linked embedding.

Area of Science:

• Computational Biology
• Genomics
• Single-Cell Analysis

Background:

• Simultaneous multi-omics measurement in single cells is advancing, yet most datasets remain unimodal.
• Integrating diverse omics data is challenging due to distinct feature spaces across modalities.

Purpose of the Study:

• To develop a computational framework, GLUE (graph-linked unified embedding), for effective integration of single-cell multi-omics data.
• To bridge the gap between different omics layers by explicitly modeling regulatory interactions.

Main Methods:

• GLUE utilizes a graph-linked approach to create a unified embedding across multiple omics layers.
• The framework models regulatory interactions to connect distinct feature spaces.

Main Results:

• GLUE demonstrated superior accuracy, robustness, and scalability compared to existing state-of-the-art tools on heterogeneous single-cell multi-omics data.
• The framework successfully performed triple-omics integration, integrative regulatory inference, and large-scale human cell atlas construction, correcting prior annotations.

Conclusions:

• GLUE provides a powerful and flexible solution for single-cell multi-omics integration and analysis.
• Its modular design allows for extension to new analytical tasks, advancing multi-omics research.