Prof.  Xuekui  Zhang
Department of Mathematics and Statistics, University of Victoria,  Canada

Title: Decoding Cellular Heterogeneity in COPD: From Single-Cell Annotation to Bulk RNA-seq Deconvolution

Abstract:

Chronic obstructive pulmonary disease (COPD) is biologically heterogeneous, but conventional bulk transcriptomic profiles obscure differences in cellular composition and cell-specific molecular states. Single-cell RNA sequencing can resolve this heterogeneity, yet translating single-cell data into clinically meaningful findings requires reliable cell-type annotation and methods that can extend cellular information to larger patient cohorts.
This talk presents a computational framework addressing these challenges. First, I will introduce a series of complementary methods for single-cell RNA-seq annotation, which illustrate the trade-offs among interpretability, computational efficiency, predictive performance, and robustness to previously unseen cell types. Using bronchoalveolar lavage datasets relevant to COPD, I will discuss how these computational tools support reproducible annotation across studies, construction of a standardized cellular atlas, and identification of robust cell-type markers, with particular attention to alveolar macrophage heterogeneity. Then, I will address a major barrier to clinical translation: single-cell RNA sequencing remains too costly and operationally demanding for many large-cohort studies. Donor-matched single-cell and bulk RNA-seq data provide a unique setting for evaluating transcriptomic deconvolution under realistic conditions. Our results show that discrepancies between synthetic pseudo-bulk references and real bulk samples can substantially impair deconvolution, whereas a modest in-study paired reference and appropriate gene filtering can materially improve existing methods. I will also discuss a fewdevolution methods being developed in my lab that explicitly use information from paired design to improve deconvolution performance.
By connecting single-cell resolution with scalable bulk profiling, this framework provides a practical foundation for studying COPD cellular heterogeneity in larger cohorts and for prioritizing candidate cellular and molecular biomarkers.

Biography: