Network-based integrative analysis of single-cell transcriptomic and epigenomic data for cell types.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell biotechnologies generate transcriptomic and epigenomic data, offering insights into cell fates.
• Integrating single-cell multi-omics data is challenging due to data heterogeneity, noise, and sparsity.
• Existing methods have limitations in effectively combining diverse single-cell omics profiles.

Purpose of the Study:

• To develop a novel network-based integrative clustering algorithm (NIC) for cell type identification.
• To fuse parallel single-cell transcriptomic (scRNA-seq) and epigenomic profiles (scATAC-seq or DNA methylation).
• To address the challenges of heterogeneity, noise, and sparsity in single-cell multi-omics data analysis.

Main Methods:

• NIC automatically learns cell-cell similarity graphs to handle data heterogeneity.
• The algorithm transforms multi-omics data fusion into the analysis of multiple networks.
• Joint non-negative matrix factorization is employed to learn shared cellular features by exploiting network structures.

Main Results:

• NIC effectively integrates scRNA-seq and epigenomic data for cell type identification.
• The algorithm demonstrated superior performance compared to state-of-the-art methods across thirteen diverse datasets.
• Experimental validation confirmed NIC's effectiveness in analyzing various tissues and organisms.

Conclusions:

• NIC provides an effective strategy for the integrative analysis of single-cell multi-omics data.
• The network-based approach overcomes key limitations in current single-cell data integration.
• The developed algorithm offers a robust tool for advancing cell type classification and understanding cellular heterogeneity.