A probabilistic gene expression barcode for annotation of cell types from single-cell RNA-seq data

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables gene expression analysis at the individual cell level, crucial for identifying and characterizing cell types.
• Accurate cell-type annotation is vital for scRNA-seq data analysis but current experimental methods are not scalable for large cell numbers.
• Existing data-driven annotation methods face limitations due to reliance on marker genes or overfitting from batch effects.

Purpose of the Study:

• To develop a robust, data-driven method for probabilistic cell-type annotation using scRNA-seq data.
• To overcome limitations of current annotation approaches, specifically marker gene dependency and batch effect sensitivity.
• To introduce a novel barcoding strategy for cell-type identification and marker gene discovery.

Main Methods:

• A statistical approach leveraging public scRNA-seq datasets.
• A latent variable model to define cell-type-specific barcodes and account for batch variations.
• Probabilistic annotation of cell identity against a reference of known cell types.

Main Results:

• The proposed barcoding approach effectively combines information across thousands of genes.
• The method successfully accounts for systematic differences (batch effects) between studies.
• Demonstrated substantial performance improvement over current reference-based annotation methods, especially in cross-study predictions.

Conclusions:

• The novel statistical approach offers a superior method for cell-type annotation in scRNA-seq data.
• The barcoding technique provides a new avenue for identifying novel marker genes.
• This method enhances the reliability and scalability of cell-type annotation for large-scale scRNA-seq studies.