Multi-modal generative modeling for joint analysis of single-cell T cell receptor and gene expression data

Area of Science:

• Immunology
• Computational Biology
• Genomics

Background:

• Single-cell immune profiling allows simultaneous transcriptome and T cell receptor (TCR) sequencing.
• Integrating multimodal immune data is challenging due to data characteristics and technical variations.

Purpose of the Study:

• To develop a multimodal generative model, mvTCR, for fusing transcriptome and TCR data.
• To demonstrate the added value of multimodal over unimodal approaches for studying immune responses.
• To enable scalable analysis of multimodal immune profiling data.

Main Methods:

• Developed the multimodal generative model mvTCR.
• Fused modality-specific information from transcriptome and TCR into a shared representation.
• Applied mvTCR to distinguish T cells binding to SARS-CoV-2 antigens.

Main Results:

• mvTCR effectively integrates transcriptome and TCR data, creating a shared representation.
• Multimodal analysis using mvTCR enhances the capture of antigen specificity compared to unimodal methods.
• mvTCR successfully distinguished T cell subpopulations binding to SARS-CoV-2 antigens.
• mvTCR facilitates mapping of new datasets to T cell references for knowledge transfer.

Conclusions:

• mvTCR offers a powerful approach for analyzing multimodal immune profiling data.
• This model advances the understanding of T cell responses and antigen specificity.
• mvTCR enables scalable analysis and knowledge transfer in immune profiling studies.