Analysis of single-cell RNA sequencing data based on autoencoders.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-Seq) is crucial for understanding development and disease.
• Effective low-dimensional data representation and dataset integration are key challenges in scRNA-Seq analysis.
• Machine learning approaches offer potential improvements for scRNA-Seq data analysis.

Purpose of the Study:

• To introduce scAEspy, a unifying tool for scRNA-Seq data analysis.
• To leverage advanced autoencoder (AE) architectures for improved data representation and integration.
• To enhance cell-type identification through effective integration of multi-platform scRNA-Seq data.

Main Methods:

• Implementation of four advanced autoencoders (AEs) and two novel AEs within the scAEspy tool.
• Integration of various batch-effect removal tools with scAEspy for multi-platform data fusion.
• Benchmarking scAEspy against existing batch-effect removal tools using scRNA-Seq datasets.

Main Results:

• scAEspy effectively captures non-linear gene interactions in scRNA-Seq data.
• Coupling scAEspy with batch-effect removal tools improves cell-type identification.
• AE-based strategies in scAEspy demonstrate superior performance compared to current solutions.

Conclusions:

• scAEspy is a user-friendly tool for analyzing scRNA-Seq data with advanced AEs.
• Its modular design allows for easy extension with new AE architectures.
• scAEspy serves as a foundation for a comprehensive toolkit for multi-omics integration.