Bayesian correlation is a robust gene similarity measure for single-cell RNA-seq data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Assessing biological information similarity is crucial for bioinformatics algorithms.
• Low expressed entities in single-cell RNA sequencing (scRNA-seq) data present challenges for accurate similarity assessment due to low read counts.
• A Bayesian correlation method, effective for bulk RNA-seq, assigns lower similarity to genes with low-confidence expression estimates.

Purpose of the Study:

• To extend the Bayesian correlation method for robust similarity assessment in scRNA-seq data.
• To evaluate Bayesian correlation's performance compared to Pearson correlation in scRNA-seq contexts.
• To investigate different strategies for computing gene similarity using Bayesian correlation.

Main Methods:

• Computed gene pair similarity across all cells.
• Identified specific cell populations and computed correlations within those populations.
• Calculated gene pair similarity across all clusters based on total mRNA expression.

Main Results:

• Bayesian correlations demonstrated higher reproducibility than Pearson correlations.
• Bayesian correlations exhibited less dependence on the number of input cells compared to Pearson correlations.
• The Bayesian correlation algorithm successfully assigned high similarity to biologically relevant genes within specific cell populations.

Conclusions:

• Bayesian correlation is a robust and reliable similarity measure for scRNA-seq data analysis.
• The method effectively addresses challenges posed by low expression levels and varying cell numbers in scRNA-seq.
• Bayesian correlation enhances the identification of biologically meaningful gene relationships in single-cell studies.