IIMPACT: Integrating Image and Molecular-Based Profiles to Analyze and Cluster Spatial Transcriptomics Data

• Activity Number: 338 - Novel Bayesian Methods in Genetic and Genomic Studies
• Type: Contributed
• Date/Time: Tuesday, August 9, 2022 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Xi Jiang* and Qiwei Li and Guanghua Xiao and Lin Xu
• Companies: Southern Methodist University and The University of Texas at Dallas and The University of Texas Southwestern Medical Center and The University of Texas Southwestern Medical Center
• Keywords: Spatial domain; Cell abundance; Markov random field; Regional spatially variable gene

Abstract

The breakthrough in spatial transcriptomics (ST) has enabled comprehensive molecular characterization at the cellular level while preserving spatial information. Meanwhile, pathology imaging powered by artificial intelligence enables the histology characterization of single cells. Understanding the spatial organization of cells and their heterogeneous gene expression profiles, will provide deeper biological insights. To address these two problems, we develop iIMPACT, a multi-stage method to cluster and analyze ST data. The first stage is an interpretable Bayesian mixture model, which combines a Gaussian component to model the molecular profile and a multinomial component for cell abundances, and incorporates the spatial information by a Markov random field prior. After region segmentation, we develop a zero-inflated generalized linear regression model under the Bayesian framework to study the association between the cellular pattern and gene expression. Applying our method to a publicly available breast cancer dataset, we found that iIMPACT outperforms existing clustering methods in terms of segmentation accuracy and generates the most biologically meaningful cancer-related genes.