Autoencoder-based cluster ensembles for single-cell RNA-seq data analysis.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables high-resolution transcriptome profiling of individual cells.
• Accurate cell type identification is crucial for scRNA-seq data analysis.
• Clustering on high-dimensional gene expression data can yield uninformative cell type clusters.

Purpose of the Study:

• To develop an improved computational framework for cell type identification using scRNA-seq data.
• To enhance the accuracy of cell clustering by addressing the challenges of high dimensionality and sparse informative genes.

Main Methods:

• Proposed an autoencoder-based cluster ensemble framework.
• Applied random subspace projections and autoencoder compression to reduce data dimensionality.
• Utilized ensemble clustering across encoded datasets to group cells.

Main Results:

• The autoencoder-based cluster ensemble significantly improved cell type-specific clustering performance.
• Performance gains up to 100% were observed compared to direct clustering on original datasets.
• The framework enhanced both k-means and SIMLR clustering algorithms.

Conclusions:

• The proposed framework facilitates more accurate cell type identification from scRNA-seq data.
• This method can improve the reliability of downstream analyses.
• The framework's code is publicly available for research use.