Exploiting single-cell expression to characterize co-expression replicability.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Co-expression networks are valuable for understanding gene function in biological processes.
• Current co-expression analysis often lacks transparency, making results difficult to interpret.
• Single-cell RNA sequencing (scRNA-seq) offers a high-resolution view of gene expression.

Purpose of the Study:

• To investigate the fundamental drivers of gene-gene connectivity and network replicability in single-cell co-expression analysis.
• To identify technical factors that influence co-expression patterns in scRNA-seq data.
• To develop strategies for controlling technical confounders in co-expression network analysis.

Main Methods:

• Analysis of co-expression networks from 31 published single-cell RNA sequencing studies.
• Performing a novel scRNA-seq experiment on 126 cortical interneurons.
• Assessing network replicability, semantic similarity, and functional connectivity.
• Re-analysis of the BrainSpan RNA sequencing dataset to confirm findings.

Main Results:

• Single-cell co-expression networks show less functional overlap with known functions compared to bulk data.
• Functional variation within cell types mirrors variation across cell types.
• Technical factors, particularly expression level, significantly impact co-expression network topology.
• Expression level is highly predictive of network structure in scRNA-seq data.

Conclusions:

• Technical properties of scRNA-seq data introduce confounds in co-expression networks.
• These confounds can be identified and controlled in supervised analyses.
• Controlling technical factors enhances co-expression analysis performance.
• This approach provides a common framework for characterizing scRNA-seq data and enabling cross-laboratory comparisons.