BiGCN Learns B Cell Functional States by Integrating Single-Cell Transcriptomes and BCR Repertoires

Area of Science:

• Immunology
• Computational Biology
• Systems Medicine

Background:

• B lymphocytes (B cells) are crucial for immune responses and disease.
• Analyzing B cell diversity is challenging due to separate transcriptomic and B cell receptor (BCR) repertoire analyses.
• Existing methods struggle to fully integrate these distinct data types for comprehensive B cell characterization.

Purpose of the Study:

• To develop a novel computational method, BiGCN, for integrated analysis of single-cell transcriptomic and BCR repertoire data.
• To enable high-fidelity characterization of B cell states and clonal relationships by unifying these modalities.
• To improve functional annotation and biological insight into B cell populations.

Main Methods:

• Developed BiGCN, a method utilizing a graph convolutional network (GCN) framework.
• Integrated paired single-cell RNA sequencing (scRNA-seq) and single-cell BCR sequencing (scBCR-seq) data.
• Applied BiGCN to diverse datasets including infectious disease, autoimmunity, and cancer.

Main Results:

• BiGCN significantly outperforms existing methods in data integration, functional annotation, and biological discovery.
• Successfully recapitulated B cell maturation and immunoglobulin class switching trajectories.
• Discovered previously unidentified intermediate B cell subtypes and functional states.

Conclusions:

• BiGCN provides a versatile and robust platform for B cell research.
• The method enhances the characterization of B cell states and clonal dynamics.
• BiGCN accelerates discoveries in immunology, cancer biology, and systems medicine.