SIMPLEs: a single-cell RNA sequencing imputation strategy preserving gene modules and cell clusters variation

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) data analysis faces challenges in reducing technical variations while preserving cell heterogeneity.
• Low mRNA content and experimental molecule losses ('dropout') lead to substantial zero read counts in scRNA-seq data.
• Current imputation methods often oversimplify gene correlations and cell type structures by treating cells or genes independently.

Purpose of the Study:

• To develop a novel statistical model-based approach for scRNA-seq data analysis.
• To address the limitations of existing imputation methods by considering gene correlations and cell type structures.
• To simultaneously impute dropouts and cluster cells while quantifying imputation and clustering uncertainty.

Main Methods:

• Proposed SIMPLEs (SIngle-cell RNA-seq iMPutation and celL clustErings), a statistical model-based approach.
• Iteratively identified correlated gene modules and cell clusters.
• Performed imputation customized for individual gene modules and cell types.
• Quantified imputation and cell clustering uncertainty using multiple imputations.

Main Results:

• SIMPLEs demonstrated superior performance compared to prevailing scRNA-seq imputation methods in simulations across various metrics.
• Application to real datasets revealed novel gene modules capable of further classifying cell subtypes.
• Imputations accurately recovered expression trends of marker genes in stem cell differentiation.
• Identified putative pathways regulating biological processes.

Conclusions:

• SIMPLEs offers an effective solution for analyzing scRNA-seq data, improving upon existing imputation techniques.
• The method enhances the identification of cell subtypes and biological pathways.
• Accurate imputation of gene expression data is crucial for understanding complex biological processes from scRNA-seq data.