Integrating feature selection with unsupervised deep embedding for clustering single-cell RNA-seq data.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution gene expression data.
• Clustering is essential for identifying distinct cell populations in scRNA-seq data.
• Current methods often perform gene selection separately, potentially missing crucial clustering information.

Purpose of the Study:

• To develop a unified framework for joint feature selection and clustering in scRNA-seq analysis.
• To address limitations of separate preprocessing steps in scRNA-seq data analysis.
• To improve the accuracy and biological relevance of cell clustering.

Main Methods:

• Proposed FSSC (Feature Selection for scRNA-seq Clustering) framework.
• Integrated a zero-inflated negative binomial (ZINB) autoencoder.
• Employed a group Lasso penalty and a dedicated clustering loss for joint optimization.

Main Results:

• FSSC simultaneously learns low-dimensional representations and selects cluster-discriminatory genes.
• The framework preserves statistical characteristics and cluster structure of scRNA-seq data.
• Consistently outperformed state-of-the-art methods on simulated and real datasets.

Conclusions:

• FSSC offers a unified approach for enhanced scRNA-seq clustering.
• The method effectively identifies biologically meaningful marker genes.
• Achieved superior clustering accuracy compared to existing methods.