Statistics Seminar: Advancing Statistical Rigor in Single-Cell and Spatial Omics Using In Silico Control Data

Presenter: Guan’ao Yan, Assistant Professor, Michigan State University

Description: Single-cell and spatial transcriptomics technologies now let us map cellular diversity and tissue organization at high resolution, but the computational methods built to analyze these data are difficult to evaluate in a rigorous, reproducible way. Two key barriers are the lack of realistic synthetic data with known ground truth and the ambiguity in how we define biologically meaningful spatial patterns. This talk will introduce two simulation frameworks—scReadSim for single-cell RNA-seq and ATAC-seq data, and scIsoSim for isoform-level expression and splicing—that generate realistic sequencing reads while preserving user-specified truth. These tools enable fair, controlled benchmarking of quantification and splicing methods across experimental protocols. The talk will also present a systematic review of 34 methods for detecting spatially variable genes (SVGs) in spatial transcriptomics data, proposing a new categorization of SVGs and outlining how future benchmarks should be designed. Overall, the goal is to improve statistical rigor, interpretability, and comparability in single-cell and spatial omics analysis.

Bio: Guan’ao Yan is an Assistant Professor of Computational Mathematics, Science & Engineering at Michigan State University. He received his Ph.D. in Statistics from UCLA. His research focuses on statistical and computational methods for modern statistical genomics, particularly single-cell and spatial omics, with an emphasis on rigorous benchmarking, interpretability, and biomedical discovery.

Hosted by: Statistics Department