An active learning approach for clustering single-cell RNA-seq data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution cellular heterogeneity data.
• Clustering is vital for identifying cell types in scRNA-seq data.
• Unsupervised clustering often yields poor biological interpretability due to separation from cell annotation.

Purpose of the Study:

• To propose an active learning (AL) framework for scRNA-seq data clustering.
• To improve the biological interpretability and efficiency of scRNA-seq data analysis.
• To develop an optimal AL approach by exploring key model parameters.

Main Methods:

• Developed an active learning (AL) framework for scRNA-seq data clustering.
• Implemented a learning algorithm that actively queries biologists for cell labels.
• Experimented with four real scRNA-seq datasets to optimize AL parameters.

Main Results:

• The proposed AL model demonstrated superior performance compared to state-of-the-art unsupervised clustering methods.
• Effective clustering was achieved with labeling of fewer than 1000 cells.
• The AL approach significantly improved biological interpretability.

Conclusions:

• Active learning is a promising tool for effective and efficient scRNA-seq data clustering.
• AL frameworks can integrate biological knowledge for more interpretable results.
• This method reduces the need for extensive manual cell labeling.