Effective Integration of Single-Cell Multi-Omics Data Using Improved Network-Based Integrative Clustering with...

Area of Science:

• Computational biology
• Genomics
• Neuroscience

Background:

• Integrating multi-omics data (snRNA-seq, snATAC-seq) is crucial for understanding cellular heterogeneity and Alzheimer's disease (AD) pathogenesis.
• Existing algorithms face challenges with high data noise, dimensionality, and computational complexity.

Purpose of the Study:

• To develop a robust algorithm for integrating glial cell multi-omics data from AD samples.
• To improve cell type identification and reveal AD pathogenesis through enhanced clustering accuracy.

Main Methods:

• Introduced multigraph regularization constraints into a network-based integrative clustering algorithm (MGR-NIC).
• Fused snRNA-seq and snATAC-seq data from AD glial cells.
• Validated MGR-NIC using simulation and real-world datasets, comparing it with NIC, scAI, MOFA v2, and JSNMF.

Main Results:

• MGR-NIC effectively removes redundant features and preserves data geometry.
• Demonstrated improved clustering accuracy and feature selection capabilities compared to the standard NIC algorithm.
• Showed strong consistency with existing clustering results on the DLPFC dataset, unlike NIC which caused cluster overlap.

Conclusions:

• MGR-NIC is a stable and reliable method for multi-omics data integration in AD research.
• The algorithm's robustness across datasets suggests its utility for accurate and consistent results.
• MGR-NIC offers a significant advancement for studying cellular heterogeneity in neurodegenerative diseases.