scSemiAAE: a semi-supervised clustering model for single-cell RNA-seq data.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) offers higher resolution of cellular diversity than bulk RNA sequencing.
• Clustering is vital for cell type identification in transcriptomics.
• Unsupervised clustering struggles with scRNA-seq's high dimensionality and dropout events, limiting biological interpretability.

Purpose of the Study:

• To develop a semi-supervised clustering model for scRNA-seq data.
• To improve the accuracy and interpretability of cell type identification.
• To leverage deep generative neural networks for enhanced clustering.

Main Methods:

• Proposed scSemiAAE, a semi-supervised clustering model.
• Utilized a Zero-Inflated Negative Binomial (ZINB) adversarial autoencoder architecture.
• Integrated adversarial training and semi-supervised modules in the latent space.

Main Results:

• scSemiAAE significantly improved clustering performance on large-scale scRNA-seq datasets.
• Outperformed numerous unsupervised and semi-supervised clustering algorithms.
• Enhanced the clustering and interpretability of downstream analyses.

Conclusions:

• scSemiAAE is an efficient Python-based tool for scRNA-seq data analysis.
• Provides effective visualization, clustering, and cell type assignment.
• The algorithm is available on GitHub for broader research application.