Visualization and Analysis of Single Cell RNA-Seq Data by Maximizing Correntropy Based Non-Negative Low Rank...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) provides a powerful lens for biological inquiry.
• Cell clustering is a key application for identifying distinct cell subtypes within scRNA-seq datasets.
• Traditional methods struggle with the inherent technical noise and data dropouts common in scRNA-seq data.

Purpose of the Study:

• To introduce a novel computational model, Maximum correntropy criterion based Non-negative and Low Rank Representation (MccNLRR), for enhanced scRNA-seq data analysis.
• To address the challenges of noise and outliers in scRNA-seq data for more accurate cell clustering.
• To leverage robust loss functions and data representation techniques for improved biological insights.

Main Methods:

• Developed the MccNLRR model incorporating the maximum correntropy criterion for noise robustness.
• Utilized low-rank representation to capture global and local data structures, including cell similarities.
• Employed an iterative algorithm based on half-quadratic optimization and alternating direction methods for optimization.
• Assessed the convergence and robustness of the MccNLRR model prior to experimental application.

Main Results:

• MccNLRR demonstrated superior performance in cell clustering tasks on scRNA-seq data.
• Visualization analyses confirmed the model's ability to delineate cell subtypes effectively.
• Gene marker selection identified key genes associated with distinct cell populations.
• The method proved accurate and robust in distinguishing cell subtypes.

Conclusions:

• The MccNLRR model offers a robust and accurate solution for analyzing scRNA-seq data.
• This approach effectively overcomes limitations of traditional methods in handling noisy and sparse single-cell data.
• MccNLRR enhances the discovery of cell subtypes and associated biological markers from scRNA-seq experiments.