Improvements Achieved by Multiple Imputation for Single-Cell RNA-Seq Data in Clustering Analysis and Differential...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) data frequently contain numerous missing values (zeroes).
• Downstream analyses like clustering and differential expression require complete datasets.
• Existing imputation methods often use single imputation, potentially introducing bias.

Purpose of the Study:

• To investigate the efficient application of multiple imputation (MI) for scRNA-seq data.
• To develop MI procedures tailored for clustering and differential expression analyses in scRNA-seq.
• To evaluate the performance of MI against single imputation for scRNA-seq data analysis.

Main Methods:

• Proposed MI procedures for clustering and differential expression analysis of scRNA-seq data.
• Utilized scIGANs (scRNA-seq imputation using Generative Adversarial Networks) to generate multiple imputed datasets.
• Applied and evaluated the MI procedures on three well-known scRNA-seq datasets.

Main Results:

• Multiple imputation (MI) demonstrated improved performance in clustering analysis.
• MI also yielded better results for differential expression analysis compared to single imputation.
• The study confirmed the advantages of MI over single imputation for scRNA-seq data.

Conclusions:

• Multiple imputation is a valuable strategy for handling missing data in scRNA-seq.
• MI enhances the accuracy and reliability of downstream analyses like clustering and differential expression.
• The proposed MI framework, integrated with scIGANs, offers a robust solution for scRNA-seq data imputation.