̽��ϵ��

Topic

SPaCeD: Spatial Point Process Distances for Pairing the Heavy and Light Chains of B Cell Receptors from Spatial BCR-seq

Department of Mathematics and Statistics

Date & location

• Thursday, March 26, 2026
• 12:00 P.M.
• Virtual Defence

Examining Committee

Supervisory Committee

• Dr. Farouk Nathoo, Department of Mathematics and Statistics, ̽��ϵ�� (Co-Supervisor)
• Dr. Brad Nelson, Department of Mathematics and Statistics, UVic (Co-Supervisor)

External Examiner

• Dr. Qihuang Zhang, Department of Epidemiology, Biostatistics and Occupational Health,
  McGill University 

Chair of Oral Examination

• Dr. Ibrahim Numanagić, Department of Computer Science, UVic

Abstract

The immune system recognizes tumor cells through antigen-specific receptors expressed by T cells and B cells. In B cells, functional receptors are formed by paired heavy and light chains, and reconstructing these pairings is essential for understanding tumor–immune interactions and cloning antigen-specific antibodies. In spatial transcriptomics experiments, heavy and light chains are often detected independently across spatial locations, leading to a challenging combinatorial pairing problem.

We propose SPaCeD, a statistical framework for inferring receptor chain pairings from spatial BCR-seq data by integrating transcriptional expression matrices with spatial co-expression point patterns. Spatial distances between clone-specific point patterns are computed using an optimal transport–based metric and combined with expression-based similarity in a unified objective function. A tuning parameter enables multiscale solutions that balance transcriptional and spatial information.

Simulation studies based on ovarian and breast cancer samples, together with real spatial B cell receptor sequencing datasets with ground truth known from single-cell sequencing, show that SPaCeD achieves higher pairing accuracy and improved stability compared with existing state-of-the-art methods. These results demonstrate that incorporating spatial structure can substantially improve receptor chain pairing in spatial transcriptomic data.

These findings highlight the potential of incorporating spatial information to improve receptor chain pairing, thereby enabling more accurate characterization of tumor–immune interactions and supporting downstream applications such as antibody discovery and immunotherapy development.