Project Summary/ Abstract Cell-cell interactions (CCI) play crucial roles in nearly all important biological processes, including cell differentiation, inflammation, wound repair and oncogenesis. CCI typically occurs when a sender cell's ligand interacts with a receiver cell's receptor, leading to changes in the target cell's transcription factor (TF) activities. Despite the advances in high-throughput methods, there is still a gap in the identification of CCIs from these data, relying largely on low-throughput manual methods. To address this gap, the proposed research aims to develop computational frameworks to model the impact of external signals on a cell's internal TF activity using a combination of single-cell multiome (scMultiome) and spatial transcriptomics (ST) data. The PI plans to create a series of computational methods based on his prior work, including: 1) DeepPrism, an improved version of BayesPrism, to more accurately deconvolve cell type fraction and cell type-specific gene expression from bulk RNA-seq and ST. 2) SpaceNiche, a method for jointly modeling the relationship between CCI and downstream gene expression using ST and provide interpretable biological insights about CCI. 3) RegulatoryVAE, a tool to infer TF activity and the relationship between regulatory elements and target genes from scMultiome which contains both the chromatin accessibility (ATAC) and gene transcription information. 4) NicheOT, a method to integrate TF activity learned from scMultiome and the spatial context learned from ST. 5) CARCell, a tool to quantify the impact of CCI on TF activity and their changes in diseased states. These tools will be able to address the significant limitations in existing methods including 1) inability to account for heterogenous gene expression and to use multiple scRNA-seq reference to perform statistical deconvolution, 2) the reliance on the transcription level of receptor/ligand to impute CCI, which may not reflect active protein levels and may fail to account for ineffective interactions due to physical separation or epigenetic states, and 3) the reliance on incomplete motif information to infer TF activity and the existing model’s inefficiency in capturing the complex relationship between DNA sequence and TF activity. The proposed methods will be applied to study the impact of CCI on inflammatory bowel disease (IBD) using scRNA-seq, scMultiome, and Visium data collected by a collaborator. IBD affects nearly 1.3% of adults in the US and its prevalence is increasing globally. Mis-regulated CCI is a key feature of the disease. By providing a generalizable tool for understanding CCI using ST and scMultiome, this work aims to fill the current gap in computational methodologies and advance the understanding of IBD, with the goal of developing new therapeutic targets.