Deep Multi-Constraint Soft Clustering Analysis for Single-Cell RNA-Seq Data via Zero-Inflated Autoencoder Embedding.

Area of Science:

• Computational Biology
• Genomics
• Data Science

Background:

• Single-cell RNA sequencing (scRNA-seq) data analysis relies on clustering cells into subgroups to understand cell heterogeneity.
• High dimensionality and sparsity of scRNA-seq data present significant challenges for accurate clustering.
• Existing methods struggle with the scale and complexity of modern scRNA-seq datasets.

Purpose of the Study:

• To propose a novel deep learning framework, single-cell Multi-Constraint deep soft K-means Clustering (scMCKC), for robust scRNA-seq data clustering.
• To address the challenges of high-dimensional and sparse scRNA-seq data.
• To improve the accuracy and performance of cell subgroup identification.

Main Methods:

• Developed scMCKC, a framework integrating a zero-inflated negative binomial (ZINB) model-based autoencoder.
• Incorporated a novel cell-level compactness constraint to enhance cluster cohesion.
• Utilized pairwise constraints from prior information and a weighted soft K-means algorithm for guided clustering.

Main Results:

• scMCKC demonstrated superior performance compared to state-of-the-art methods across eleven scRNA-seq datasets.
• The method showed robust and excellent performance in clustering analysis, validated on a human kidney dataset.
• Ablation studies confirmed the effectiveness of the novel cell-level compactness constraint.

Conclusions:

• scMCKC provides a powerful and effective approach for scRNA-seq data clustering.
• The proposed framework successfully addresses the challenges of data sparsity and high dimensionality.
• The novel constraints significantly contribute to improved clustering accuracy and biological insights.