TsImpute: an accurate two-step imputation method for single-cell RNA-seq data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution gene expression data.
• Technical limitations cause "dropouts" (excessive zeros) in scRNA-seq, potentially distorting biological insights.
• Accurate imputation of dropouts is essential for reliable downstream analysis.

Purpose of the Study:

• To develop and evaluate tsImpute, a novel two-step imputation method for scRNA-seq data.
• To effectively distinguish true zeros from technical dropouts.
• To improve the accuracy of gene expression recovery, cell clustering, and differential expression analysis.

Main Methods:

• tsImpute employs a two-step imputation strategy.
• Step 1: Utilizes a zero-inflated negative binomial distribution to identify dropouts and perform initial imputation.
• Step 2: Conducts cell clustering on the modified expression matrix, followed by distance-weighted imputation.

Main Results:

• tsImpute demonstrates superior performance in recovering gene expression compared to existing methods.
• The method enhances the accuracy of cell clustering.
• It also improves the reliability of differential expression analysis.
• Evaluated using both simulated and real scRNA-seq datasets.

Conclusions:

• tsImpute is an effective computational tool for addressing dropout issues in scRNA-seq data.
• The method improves the biological interpretability of scRNA-seq datasets.
• The R package for tsImpute is publicly available for researchers.