scMIC: A Deep Multi-Level Information Fusion Framework for Clustering Single-Cell Multi-Omics Data.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Cell type identification is essential for understanding cellular heterogeneity and biological processes.
• Single-cell sequencing technologies have advanced cell clustering methods, primarily for single omics data.
• Integrating single-cell multi-omics data presents computational challenges for existing methods.

Purpose of the Study:

• To propose a novel deep multi-level information fusion framework, scMIC, for accurate cell clustering using multi-omics data.
• To address the challenge of integrating consensus and complementary information from multiple omics for improved cell type identification.
• To enhance the robustness and accuracy of cell clustering by leveraging multi-omics data.

Main Methods:

• Developed scMIC, a deep multi-level information fusion framework for single-cell multi-omics data.
• Integrated cell attribute information and structural relationships at local and global levels.
• Employed a multiple collaborative supervised clustering strategy to guide representation learning and omics information exchange.

Main Results:

• scMIC effectively integrates multi-omics data by reducing redundancy and enhancing discriminative representations.
• The framework leverages both consensus and complementary information across different omics.
• Experimental results on seven datasets demonstrate scMIC's superior performance compared to state-of-the-art methods.

Conclusions:

• scMIC provides a robust and accurate approach for cell type identification using single-cell multi-omics data.
• The deep multi-level fusion framework effectively addresses the challenges of integrating diverse omics information.
• This method advances the field of computational biology by improving cell clustering accuracy.