scSCC: A swapped contrastive learning-based clustering method for single-cell gene expression data.

Area of Science:

• Computational Biology
• Genomics
• Data Science

Background:

• Cell clustering is crucial for identifying cell types in single-cell RNA sequencing (scRNA-seq) data.
• Accurate cell clustering facilitates downstream cell annotation and biological interpretation.

Purpose of the Study:

• To develop a novel contrastive clustering algorithm for scRNA-seq data.
• To improve the accuracy and robustness of cell clustering in scRNA-seq analysis.

Main Methods:

• Proposed scSCC, a swapped contrastive clustering algorithm for scRNA-seq data.
• Integrated instance contrastive learning and a swapped prediction module to learn disentangled cell representations.
• Utilized clustering prototypes within the swapped prediction module to inject clustering signals.

Main Results:

• scSCC demonstrated improved clustering performance on real scRNA-seq datasets compared to existing methods.
• Ablation studies confirmed the effectiveness of combining instance learning and the swapped prediction module.
• The algorithm generates more clustering-friendly cell representations by encouraging convergence to prototypes.

Conclusions:

• scSCC offers a powerful new approach for cell clustering in scRNA-seq data analysis.
• The combination of contrastive learning modules effectively enhances clustering signals.
• The proposed method provides more accurate and interpretable cell type identification.