MuDCoD: multi-subject community detection in personalized dynamic gene networks from single-cell RNA sequencing

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Population-scale single-cell RNA sequencing (scRNA-seq) data enable novel biological insights.
• Existing methods for gene co-expression network analysis do not simultaneously account for multiple subjects and time points.
• Understanding subject- and time-specific variations in gene networks is crucial for explaining phenotypic differences.

Purpose of the Study:

• To develop a novel computational framework for multi-subject community detection in personalized dynamic gene networks from scRNA-seq data.
• To identify gene communities that vary or are shared across time and subjects.

Main Methods:

• Developed Multi-subject Dynamic Community Detection (MuDCoD), a method based on spectral clustering.
• MuDCoD promotes information sharing across networks from different subjects and time points.
• Applied MuDCoD to analyze scRNA-seq datasets of human-induced pluripotent stem cells and CD4+ T cells.

Main Results:

• MuDCoD effectively leverages shared signals among subject-specific networks and performs robustly with limited information sharing.
• The method successfully identified time-varying personalized gene modules in real-world scRNA-seq datasets.
• Demonstrated the utility of personalized dynamic community detection for exploring subject-specific biological processes.

Conclusions:

• MuDCoD provides a robust approach for analyzing complex, multi-subject, longitudinal scRNA-seq data.
• The framework enables the discovery of dynamic, personalized gene regulatory patterns.
• Personalized dynamic community detection can significantly advance the understanding of individual variability in biological systems.