Single-cell RNA-seq data semi-supervised clustering and annotation via structural regularized domain adaptation.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) enables exploration of tissue heterogeneity.
• Accurate cell type identification is crucial for scRNA-seq data analysis and understanding heterogeneity.
• Traditional cell annotation methods are labor-intensive and resource-demanding as data scales increase.

Purpose of the Study:

• To develop a flexible semi-supervised framework for single-cell clustering and annotation.
• To address limitations of existing supervised methods, such as batch effects and compromised target data discrimination.
• To improve the efficiency and accuracy of cell type identification in scRNA-seq data.

Main Methods:

• Propose scSemiCluster, a semi-supervised framework inspired by unsupervised domain adaptation.
• Integrate reference and target scRNA-seq data for model training.
• Utilize structure similarity regularization and pairwise constraints in feature learning.

Main Results:

• scSemiCluster outperforms state-of-the-art supervised and semi-supervised methods on both simulated and real scRNA-seq data.
• The framework achieves high accuracy without explicit domain alignment or batch effect correction.
• Demonstrates the effectiveness of integrating deep discriminative and generative clustering techniques.

Conclusions:

• scSemiCluster provides an effective solution for cell type annotation in scRNA-seq data.
• The proposed method enhances accuracy and robustness against batch effects.
• This work pioneers the application of combined deep clustering techniques in single-cell analysis.