Comparison of Computational Methods for Imputing Single-Cell RNA-Sequencing Data.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) enables high-resolution gene expression analysis.
• Dropout events (excessive zeros) are a major challenge in scRNA-seq data, potentially distorting downstream analyses.
• Various imputation methods exist to address dropout effects, but their performance and characteristics vary significantly.

Purpose of the Study:

• To conduct a large-scale comparison and evaluation of existing gene expression imputation methods for scRNA-seq data.
• To assess the impact of imputation methods on critical downstream analyses, including cell type clustering, differential gene expression detection, and lineage trajectory reconstruction.

Main Methods:

• Compared eight different gene expression imputation methods.
• Evaluated method performance using both simulated and real scRNA-seq datasets.
• Assessed the influence of imputation on cell clustering, differential expression analysis, and lineage tracing.

Main Results:

• No single imputation method demonstrated superior performance across all tested scenarios and datasets.
• The choice of imputation method can significantly affect the outcomes of downstream analyses.
• Identified limitations including scalability, robustness, and situational availability across the evaluated methods.

Conclusions:

• A comprehensive, large-scale evaluation is crucial due to the diverse performance of scRNA-seq imputation methods.
• Future research should focus on addressing the identified defects to develop more robust and versatile imputation tools.
• Careful consideration and selection of imputation methods are necessary for reliable scRNA-seq data interpretation.