Detection of high variability in gene expression from single-cell RNA-seq profiling.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) enables detailed analysis of cell heterogeneity.
• Identifying highly variable genes is crucial for scRNA-seq data analysis.
• Existing methods lack a universal model for gene expression variation across diverse scRNA-seq datasets.

Purpose of the Study:

• Develop a generic gene expression variation model (GEVM) for scRNA-seq data.
• Quantify variably expressed genes (VEGs) by addressing data over-dispersion.
• Establish a statistically robust method for analyzing gene expression variability.

Main Methods:

• Utilized the relationship between coefficient of variation (CV) and average expression level.
• Developed a simulation framework to generate and test scRNA-seq data.
• Evaluated model robustness using root-mean-square error (RMSE) and real scRNA-seq datasets.

Main Results:

• Demonstrated robust parameter estimation for the GEVM using simulated data.
• Achieved minimal root mean square errors in regression analysis.
• Successfully identified VEGs in distinct real scRNA-seq datasets, revealing cell subpopulations.

Conclusions:

• The proposed GEVM effectively identifies significant variably expressed genes.
• The model performs robustly across different scRNA-seq datasets and protocols.
• GEVM facilitates the discovery of cell heterogeneity and subpopulations.