时 间：2026年 9月22日（周二）15:00-16:00
地 点：普陀校区理科大楼A1514室
报告人：刘丙强 山东大学数学学院教授
主持人：於州 华东师范大学统计学院
摘 要:
Multicellular organisms rely on the cooperation between cells to carry out complex biological processes and the spatial organization of cells is crucial for their coordinated function. Spatial transcriptomics technologies now enable high-throughput profiling of gene expression within tissue sections while preserving the inherent spatial context. The inherent complexity of such data has driven the rapid evolution of advanced computational frameworks designed to decipher the underlying logic of biological systems. By modeling spatial context as a graph and viewing cellular molecular profiles as graph signals, graph signal processing (GSP) theory can be leveraged to characterize the spatial patterns of gene expression across complex tissue architectures. Here, we present a suite of computational methods based on GSP and deep learning to address diverse analytical tasks in spatial transcriptomics, including gene spatial representation (SpaGFT), spatial domain identification (DeepGFT), cell-cell communication (SpaGRD), molecular lateral diffusion recovery (DiffGSP), cell-type deconvolution (STAID), and cell imputation (CellBack). These methods not only provide effective solutions for specific analytical challenges in spatial omics but also extend the reach of mathematical modeling and theoretical frameworks to the broader, rapidly evolving field of spatial biology.
报告人简介：
刘丙强，山东大学数学学院教授、博士生导师。主要从事数学与生物医学交叉研究。获批主持国家重点研发计划、基金委青年基金A类等科研项目。担任中国运筹学会理事、中国工业与应用数学学会数学生命科学专委会委员、中国数学会生物数学专委会委员、中国计算机学会生物信息学专委会委员、山东数学会副秘书长、山东省生物信息学会副理事长、Computational Biology and Chemistry副主编等学术职务。