A posterior probability based Bayesian method for single-cell RNA-seq data imputation.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) data are characterized by a high proportion of zero counts, known as dropout events.
• These dropout events significantly hinder accurate downstream data analysis and interpretation.

Purpose of the Study:

• To introduce BayesImpute, a novel method for inferring and imputing dropout events in scRNA-seq data.
• To evaluate the performance of BayesImpute in identifying dropouts, recovering true expression levels, and preserving biological information.

Main Methods:

• BayesImpute utilizes gene expression rates and coefficients of variation within cell subpopulations to identify potential dropouts.
• It constructs posterior distributions for each gene and employs the posterior mean for imputation.
• The method was validated using simulated and real scRNA-seq datasets.

Main Results:

• BayesImpute accurately identifies dropout events and minimizes the introduction of false positive signals.
• The method successfully recovers true expression levels of missing values and restores gene-gene and cell-cell correlations.
• BayesImpute enhances cell subpopulation clustering, visualization, and the identification of differentially expressed genes.
• Compared to other methods, BayesImpute demonstrates scalability, speed, and minimal memory usage.

Conclusions:

• BayesImpute is an effective and efficient tool for handling dropout events in scRNA-seq data.
• The method improves the reliability and biological relevance of scRNA-seq data analysis.
• BayesImpute offers a scalable and fast imputation solution for large-scale single-cell studies.