Sha Cao
Indiana University School of Medicine
3:30 PM - 4:30 PM
Zoom Link: https://notredame.zoom.us/j/92712074286?pwd=ZzZGc1pzMXczalhtQXZONVkxcXpXQT09
Meeting ID: 927 1207 4286
Passcode: 952926
Title: Modeling Single Cell Transcriptomics Data by Graph Structure Embedding
Abstract: The development of single cell transcriptomics technology has provided gene expressions at the resolution of single cells, and are promising to deliver better understanding of the complexity within the disease microenvironment. Many pure data-driven models for single cell expression data have been developed, but models that incorporate known graph structures are much less abundant. Popular graph structures in biomedical knowledge database include the whole metabolic network and biological pathway networks, both of which are set of metabolic and biological processes that determine the physiological and biochemical properties of a cell. We here are presenting two models for single cell expression data: the first model, based on graph neural network, aims to incorporate the whole metabolic network to study the metabolic states of single cells, as well as the metabolic heterogeneity, and exchanges among different cell groups; the second model, based on covariance regression, aims to incorporate the biological pathway information to study the variations of gene-gene interactions, which allow rigorous statistical inferences for important covariates, such as disease status, sex, etc. Both models have been applied to important disease types, including cancer and Alzheimer’s Disease.