Improving the performance of single-cell RNA-seq data mining based on relative expression orderings

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) offers deep insights into cellular heterogeneity.
• Gene expression variability within single cells necessitates more robust feature representations.
• Existing methods struggle to capture stable, reliable features for scRNA-seq data analysis.

Purpose of the Study:

• To introduce a novel feature matrix, the delta rank matrix (DRM), for scRNA-seq data.
• To transform unreliable gene expression values into stable gene interaction/edge values at the single-cell level.
• To enhance the analysis of cell heterogeneity, identification, and relationships using integrated gene expression and interaction data.

Main Methods:

• Constructed the delta rank matrix (DRM) by integrating scRNA-seq data with a priori gene interaction networks.
• Utilized relative gene expression orderings to derive stable interaction/edge-level features.
• Applied DRM to various scRNA-seq datasets for comparative analysis.

Main Results:

• DRM demonstrated superior performance in cell clustering, identification, and pseudo-trajectory reconstruction compared to the original gene expression matrix.
• Successfully fused gene expression and gene interaction information, enabling single-cell level measurement of gene interactions.
• Facilitated the identification of changes in gene interactions across different cell types.

Conclusions:

• DRM provides a stable and reliable feature matrix for scRNA-seq data analysis, overcoming limitations of raw gene expression values.
• Enables a novel perspective for scRNA-seq analysis by focusing on gene interactions at the single-cell level.
• Offers a new approach for constructing cell-specific networks, advancing our understanding of biological systems.