The advent of spatial transcriptomics has greatly expanded the experimental toolbox to study cellular interactions and tissue architecture, but existing methods are primarily focused on identifying communications at the cell-type level and often overlook the intricate dynamics between individual cells. We introduce IC3, a novel probabilistic graphical model that infers cell-cell communications at the single-cell level using spatial transcriptomics data. Each cell is separately represented in the network, allowing heterogeneity for cells of the same cell type. Moreover, we devise a scan framework to identify communication hotspots, which are local regions with high concentrations of cell-cell communications, explicitly showcasing how the finer resolution network provides deeper biological insight for tissue architecture and function.

侯琳, 清华大学统计学研究中心长聘副教授、博士生导师, 清华大学统计与数据科学系副系主任。主要从事生物统计、生物信息、统计遗传学等方向的研究。她于2011年获得北京大学统计学博士学位, 2012年至2015年在耶鲁大学生物统计系从事研究工作, 历任博士后、副研究员, 2015年起加入清华大学统计学研究中心。担任中国现场统计研究会计算统计分会常务理事、秘书长; Statistics in Biosciences编委，Quantitative Biology编委。