transCAE: Enhancing Cell Type Annotation in Single-cell RNA-seq Data with Transfer Learning and Convolutional...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for biological research.
• Accurate cell type annotation using reference datasets is a key application.
• Existing methods face challenges with data quality and batch effects.

Purpose of the Study:

• To develop a robust algorithm for precise scRNA-seq data annotation.
• To overcome limitations of existing supervised and semi-supervised methods.
• To effectively mitigate batch effects in multi-dataset analyses.

Main Methods:

• Developed transCAE, a transfer learning-based algorithm.
• Integrated unsupervised dimensionality reduction with supervised classification.
• Leveraged information from both reference and query datasets.

Main Results:

• transCAE significantly enhances cell classification accuracy.
• The algorithm efficiently mitigates batch effects.
• Demonstrated superior performance compared to state-of-the-art methods in multi-dataset experiments.

Conclusions:

• transCAE offers a robust and optimal solution for scRNA-seq annotation.
• The method effectively addresses challenges posed by data quality and batch effects.
• Positions transCAE as a leading tool for scRNA-seq data analysis.