DEsingle

DEsingle models zero inflation in single-cell RNA sequencing (scRNA-seq) data using a Zero-Inflated Negative Binomial framework to distinguish real and dropout zeros and detect differential expression between two cell groups.

Key Features:

• Zero-Inflated Negative Binomial (ZINB) modeling: Uses the ZINB model to account for excess zeros in scRNA-seq data.
• Real versus dropout zero estimation: Estimates the proportion of real zeros versus dropout zeros to separate biological absence of expression from technical dropouts.
• DE gene classification: Identifies three categories of differentially expressed genes between two groups: Differential Expression Status (DEs), Differential Expression Abundance (DEa), and General Differential Expression (DEg).
• Detection improvement: Enhances accuracy of differential expression detection by explicitly modeling zero inflation in scRNA-seq data.
• Implementation: Implemented as an R package and references Bioconductor and supplementary data reported in Bioinformatics online.

Scientific Applications:

• Differential expression analysis in scRNA-seq: Detects genes with differences in expression abundance or transcriptional on/off states between two cell groups in single-cell RNA-seq datasets.
• Characterizing transcriptional states: Distinguishes genes with changes in on-off transcriptional status (DEs) from those with changes in expression abundance (DEa).
• Studying cellular heterogeneity and dynamics: Aids investigation of cellular heterogeneity and transcriptional dynamics by separating biological zeros from technical dropouts.

Methodology:

Applies the Zero-Inflated Negative Binomial (ZINB) statistical model to estimate proportions of real versus dropout zeros and classifies genes into DEs, DEa, and DEg between two cell groups.