Determining gene specificity from multivariate single-cell RNA sequencing data

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Identifying genes specific to biological categories or experimental conditions is crucial in single-cell genomics.
• Existing methods for gene specificity analysis lack a standardized axiomatic approach.

Purpose of the Study:

• To develop a robust and axiomatic method for measuring gene specificity.
• To apply this method to uncover novel gene expression patterns in mouse and human datasets.

Main Methods:

• Developed ember (Entropy Metrics for Biological ExploRation), an axiomatic approach for gene specificity measurement.
• Applied ember to single-cell RNA sequencing data from eight mouse tissues and founder strains.
• Utilized unsupervised learning on entropy metrics to identify specialized genes.
• Extended ember analysis to a human peripheral blood mononuclear cell (PBMC) dataset.

Main Results:

• Ember is the only method satisfying four key properties for gene specificity.
• Mouse gene specificity is often unintuitive, with context-dependent housekeeping genes and strain-driven cell type switching.
• Uncovered shared genes specific to male gonads and kidneys, and stage-specific kidney genes.
• Human PBMC variation is concentrated in classical monocytes, with sex, age, and ancestry-specific genes identified.

Conclusions:

• Ember provides a powerful, axiomatic tool for gene specificity analysis in single-cell genomics.
• The study reveals complex and context-dependent gene expression patterns in mice and humans.
• Findings offer insights into genetic variation and provide a roadmap for future research using the murine model.