WEDGE: imputation of gene expression values from single-cell RNA-seq datasets using biased matrix decomposition

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell sequencing technologies offer high resolution but suffer from low RNA capture rates, leading to sparse expression matrices.
• This sparsity poses a significant challenge for downstream functional genomics analyses and accurate biological interpretation.
• Existing imputation methods struggle to effectively recover missing values in highly sparse single-cell transcriptome data.

Purpose of the Study:

• To develop a novel algorithm for imputing gene expression matrices from single-cell RNA sequencing data.
• To address the limitations of current methods in handling sparse expression data.
• To enhance the accuracy and utility of single-cell RNA sequencing datasets for biological discovery.

Main Methods:

• Proposed a new algorithm named WEDGE (WEighted Decomposition of Gene Expression).
• Employed a biased low-rank matrix decomposition method for imputation.
• Applied WEDGE to sparse gene expression matrices.

Main Results:

• WEDGE successfully recovered gene expression matrices.
• The algorithm accurately reproduced cell-wise and gene-wise correlations.
• WEDGE demonstrated improved cell clustering performance, especially on sparse datasets.

Conclusions:

• WEDGE provides a potent approach for imputing sparse expression matrix data.
• The WEDGE algorithm can help researchers extract more biological meaning from single-cell RNA sequencing data.
• The source code for WEDGE is publicly available for broader research application.