Single-cell gene set enrichment analysis and transfer learning for functional annotation of scRNA-seq data

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Cell functional annotation is crucial but challenging for single-cell transcriptional data.
• Existing methods often adapt bulk RNA sequencing techniques or rely on marker genes from clustering.
• These approaches can be limited and lack automation for accurate cell function determination.

Purpose of the Study:

• To develop novel, automated methods for cell functional annotation using single-cell RNA sequencing (scRNA-seq) data.
• To overcome limitations of existing annotation techniques by integrating advanced computational approaches.
• To provide a robust framework for improved interpretation of scRNA-seq datasets.

Main Methods:

• Single-cell gene set enrichment analysis (scGSEA): Combines latent data representations and gene set enrichment scores for coordinated gene activity detection.
• Single-cell mapper (scMAP): Employs transfer learning to contextualize new cells within a reference cell atlas.
• Validation using simulated and real scRNA-seq datasets.

Main Results:

• scGSEA effectively identifies conserved pathway activity patterns across different experimental conditions at single-cell resolution.
• scMAP accurately maps and contextualizes new single-cell profiles onto a reference breast cancer atlas.
• Both methods demonstrate high performance in recapitulating and assigning cell functions.

Conclusions:

• scGSEA and scMAP offer a powerful, automated workflow for cell functional annotation.
• These novel tools significantly enhance the annotation and interpretation of scRNA-seq data.
• The developed framework provides a valuable resource for single-cell data analysis and biological discovery.