ScLRTC: imputation for single-cell RNA-seq data via low-rank tensor completion.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) reveals gene expression at high resolution.
• Technical limitations cause dropout events, leading to missing data and noise in scRNA-seq matrices.
• Accurate gene expression recovery is crucial for reliable downstream analyses.

Purpose of the Study:

• To propose a novel method, scLRTC, for imputing dropout events in scRNA-seq data.
• To enhance the accuracy of gene expression matrices by recovering true expression levels.
• To improve downstream analyses, including cell classification, visualization, and lineage trajectory inference.

Main Methods:

• Developed scLRTC, a low-rank tensor completion-based method.
• Utilized cell similarity to construct a third-order low-rank tensor.
• Employed tensor decomposition for data denoising and low-rank tensor completion for reconstruction.
• Restored gene-to-gene and cell-to-cell correlations.

Main Results:

• scLRTC outperformed state-of-the-art methods on simulated datasets, assessed by SSE and PCC.
• Achieved superior cell classification accuracy on real datasets, evaluated by ARI and NMI.
• Demonstrated effectiveness in cell visualization and inferring cell lineage trajectories.

Conclusions:

• scLRTC offers improved imputation results compared to existing tools.
• The method enhances the accuracy and reliability of scRNA-seq data analysis.
• Source code is publicly available for reproducibility and further research.