SEAL: Semantic-Aware Contrastive Learning for scRNA-Seq Clustering.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) allows cellular-level biological exploration.
• Unsupervised clustering is crucial for identifying cell types in scRNA-seq data.
• Existing clustering methods struggle with unstable performance due to data noise and high dropout rates.

Purpose of the Study:

• To develop a novel clustering approach for scRNA-seq data that overcomes limitations of current methods.
• To enhance the accuracy and stability of cell type identification in scRNA-seq analysis.
• To leverage semantic information for improved representation learning in scRNA-seq data.

Main Methods:

• Proposed a SEmantic-Aware contrastive Learning (SEAL) framework for scRNA-seq clustering.
• Generated data augmentations by randomly masking gene expression in each cell.
• Applied semantic-aware contrastive learning using pseudo-labels to capture invariant representations.

Main Results:

• SEAL effectively learns biologically meaningful cell representations.
• The proposed method demonstrates accurate cell type identification.
• SEAL shows improved performance in handling high dropout rates and noise inherent in scRNA-seq data.

Conclusions:

• SEAL offers a robust and effective solution for unsupervised clustering of scRNA-seq data.
• The semantic-aware contrastive learning approach enhances the interpretability and accuracy of cell type discovery.
• This method advances scRNA-seq data analysis by providing more stable and reliable clustering results.