baredSC: Bayesian approach to retrieve expression distribution of single-cell data.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) is increasingly used to study cellular transcriptomes.
• Data sparsity and technical noise, particularly Poisson noise from sampling, pose challenges in scRNA-seq analysis.
• Accurate gene expression distribution and correlation analysis are crucial for understanding cellular heterogeneity.

Purpose of the Study:

• To introduce baredSC, a novel computational tool for inferring intrinsic gene expression distributions from sparse scRNA-seq data.
• To address the limitations of existing methods in handling sampling noise and data sparsity.
• To enable robust analysis of gene expression in one and two dimensions.

Main Methods:

• Development of baredSC, employing a Bayesian approach with a Gaussian mixture model.
• Application to simulated scRNA-seq data to assess performance on sparse and multi-modal distributions.
• Validation on real biological datasets, including gene-gene correlations and complex expression patterns.

Main Results:

• baredSC successfully infers underlying gene expression distributions, even with highly sparse and multi-modal data.
• The tool accurately estimates gene expression correlations, demonstrating robustness against sampling noise.
• baredSC identified a trimodal expression distribution for Pitx1 in embryonic hindlimb, consistent with flow cytometry data.

Conclusions:

• baredSC is an effective tool for retrieving gene expression distributions from scRNA-seq data.
• The method enhances the analysis of gene expression patterns and correlations in challenging datasets.
• baredSC offers a significant advancement for researchers working with scRNA-seq data.