时 间：2024年10月18日13:00-14:00
地 点：普陀校区理科大楼A1514
报告人：孔德含 加拿大多伦多大学副教授
主持人：项冬冬 华东师范大学教授
摘 要：
In this presentation, I will discuss two applications of optimal transport in the context of single-cell RNA sequencing. In the first part, we introduce Laplacian Linear Optimal Transport (LLOT), a biologically interpretable method designed to integrate single-cell and spatial transcriptomics data, enabling the reconstruction of missing information at both whole-genome and single-cell resolution. LLOT has two key features: it efficiently identifies differences between datasets and corrects platform effects through a linear mapping approach, and it adeptly manages complex spatial structures within tissues. We benchmarked LLOT against several alternative methods using real datasets, and the results consistently demonstrated its superior performance in predicting spatial gene expression and single-cell locations. In the second part, we develop a novel method based on discrete unbalanced optimal transport to model cell type developmental trajectories. This method detects biological changes in cell types and infers transitions to various states using the transport matrix. We evaluated it with single-cell RNA data from mouse embryonic fibroblasts, where it accurately identified major cell type developmental changes, validated by experimental results. Additionally, the transition probabilities between cell types revealed a high level of biological precision.
报告人简介：
孔德含，多伦多大学统计学副教授，研究方向包括脑图像，统计遗传和基因组学，函数型数据分析，因果推断，高维数据分析以及机器学习。研究成果发表在统计学国际顶级期刊JRSSB，JASA，Biometrika等，现任统计学期刊JASA副主编。