Entropy subspace separation-based clustering for noise reduction (ENCORE) of scRNA-seq data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for understanding cellular heterogeneity.
• Data noise in scRNA-seq compromises the accuracy of clustering, marker identification, and visualization.
• Existing methods struggle to effectively distinguish true biological signals from noise.

Purpose of the Study:

• To develop a robust cell clustering algorithm that effectively reduces noise in scRNA-seq data.
• To improve the accuracy of cell type identification and marker discovery from noisy scRNA-seq datasets.
• To provide a high-resolution visualization tool for scRNA-seq data analysis.

Main Methods:

• Proposed 'entropy subspace' separation strategy to identify informative features by analyzing feature density profiles.
• Developed ENtropy subspace separation-based Clustering for nOise REduction (ENCORE) algorithm.
• Integrated 'entropy subspace' separation with a consensus clustering method.

Main Results:

• ENCORE demonstrated superior performance in cell clustering across 12 standard scRNA-seq datasets.
• Achieved high-resolution visualization of cellular populations.
• Successfully identified biologically significant group markers, even from challenging datasets.
• Showcased effective feature selection capabilities.

Conclusions:

• ENCORE is an effective tool for scRNA-seq data analysis, offering improved clustering, accurate marker identification, and high-resolution visualization.
• The 'entropy subspace' separation strategy effectively distinguishes informative features from noise.
• ENCORE facilitates the study of cellular heterogeneity and the discovery of group markers.