A novel method for single-cell data imputation using subspace regression.

Area of Science:

• Biochemistry
• Genomics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) enables high-resolution biological system analysis.
• Low mRNA capture in scRNA-seq leads to missing expression values, known as dropouts, hindering accurate quantification.
• Dropouts are a significant challenge in interpreting scRNA-seq data.

Purpose of the Study:

• To develop a novel imputation method, single-cell Imputation via Subspace Regression (scISR), for accurate recovery of dropout values in scRNA-seq data.
• To address the challenge of missing data in single-cell gene expression profiles.

Main Methods:

• scISR employs a hypothesis-testing technique to identify dropout-affected zero-valued entries.
• Dropout values are estimated using a subspace regression model.
• The method was evaluated on 25 scRNA-seq datasets and simulation scenarios.

Main Results:

• scISR demonstrated superior performance in recovering scRNA-seq expression profiles compared to five state-of-the-art imputation methods.
• The method consistently enhanced the quality of cluster analysis across various dropout rates, normalization techniques, and quantification schemes.
• scISR effectively addresses dropout events, leading to more reliable gene expression data.

Conclusions:

• scISR provides a robust solution for imputing missing values in scRNA-seq data.
• The developed method improves the accuracy and reliability of downstream analyses, such as clustering.
• scISR offers a valuable tool for researchers working with single-cell gene expression data.