scGMAI: a Gaussian mixture model for clustering single-cell RNA-Seq data based on deep autoencoder.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-Seq) enables detailed gene expression analysis but faces challenges like dropout events and high dimensionality.
• Existing computational tools for scRNA-Seq data analysis have limitations in accurately clustering cell types.

Purpose of the Study:

• To develop and validate scGMAI, a new computational method for accurate cell clustering in scRNA-Seq data.
• To overcome common obstacles in scRNA-Seq data analysis, such as dropout events and dimensionality.

Main Methods:

• scGMAI employs autoencoder networks for reconstructing gene expression values from scRNA-Seq data.
• Fast independent component analysis (FastICA) is utilized for dimensionality reduction of the reconstructed data.
• A Gaussian mixture model is integrated for robust cell clustering.

Main Results:

• scGMAI demonstrated superior performance in cell clustering across 17 public scRNA-Seq datasets compared to established methods like Seurat.
• The method effectively handles noise and reduces dimensionality, leading to more accurate cell type identification.
• Validation on diverse datasets confirms the robustness and accuracy of scGMAI.

Conclusions:

• scGMAI is an effective and accurate tool for clustering and identifying cell types from scRNA-Seq data.
• The integration of autoencoders and FastICA offers a powerful approach to overcoming scRNA-Seq analysis challenges.
• scGMAI shows significant potential for broad applications in single-cell genomics research.