scLRTC

scLRTC applies low-rank tensor completion to impute dropout events in single-cell RNA sequencing (scRNA-seq) data, restoring gene–cell expression matrices and preserving gene-to-gene and cell-to-cell correlations for downstream analyses.

Key Features:

• Imputation Accuracy: Imputes dropout entries and quantifies accuracy using sum of squared error (SSE) and Pearson correlation coefficient (PCC).
• Performance on Simulated Datasets: Outperforms other state-of-the-art methods on simulated scRNA-seq datasets in imputation fidelity.
• Effectiveness on Real Datasets: Improves cell classification on real scRNA-seq datasets across clustering approaches such as SC3 or t-SNE followed by K-means, evaluated by adjusted rand index (ARI) and normalized mutual information (NMI).
• Correlation Preservation: Restores both gene-to-gene and cell-to-cell correlations through reconstruction of expression values.
• Versatility in Applications: Supports downstream tasks including cell visualization and inference of cell lineage trajectories.

Scientific Applications:

• Cell Type Clustering: Provides an imputed expression matrix as input for cell type clustering analyses.
• Dimension Reduction and Visualization: Reduces dropout-induced noise to improve dimension reduction and visualization methods such as t-SNE.
• Cell Lineage Trajectory Inference: Supplies denoised expression data for inferring cell lineage trajectories.

Methodology:

Constructs a third-order low-rank tensor leveraging similarities among single cells, applies tensor decomposition to denoise the data, and uses a low-rank tensor completion algorithm to reconstruct cell expression levels.