Fractional Order Total Variation Low-Rank Representation on Single-Cell RNA Sequencing Clustering.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Bulk RNA sequencing masks crucial cell-to-cell variability.
• Single-cell RNA sequencing (scRNA-seq) captures cellular heterogeneity but faces challenges with high dimensionality, noise, and dropout events.
• Accurate clustering and subtype identification are vital for understanding cell development, differentiation, and disease.

Purpose of the Study:

• To develop an advanced computational framework for robust single-cell clustering.
• To address the challenges of noise and data sparsity in scRNA-seq data.
• To improve the accuracy and reliability of cell type identification from scRNA-seq data.

Main Methods:

• Proposed an Adaptive Fractional-Order Total Variation Regularised Low-Rank Representation (AFTV-LRR) model.
• Integrated adaptive fractional-order total variation with low-rank representation for feature preservation.
• Employed the Alternating Direction Method of Multipliers (ADMM) for optimization and spectral clustering for cell identification.

Main Results:

• AFTV-LRR demonstrated competitive or superior performance against eight existing algorithms on 11 scRNA-seq datasets, measured by Adjusted Rand Index (ARI) and Normalised Mutual Information (NMI).
• t-SNE visualizations confirmed clearer inter-cluster separations and higher intra-cluster compactness.
• Marker gene analysis on a mouse embryo dataset validated the biological interpretability and robustness of the clustering results.

Conclusions:

• The proposed AFTV-LRR model offers an adaptive computational framework for enhanced single-cell clustering.
• This method effectively reconstructs low-rank structures while preserving fine-grained cellular features.
• The findings contribute to more accurate and reliable analyses of scRNA-seq data for biological discovery.