UC San Diego

Cell–cell interactions and communication govern organismal development, homeostasis, and cellular functions. Recent advances in RNA sequencing technologies and the growing knowledge about ligand–receptor interactions allow for routine analysis of intercellular signaling from gene expression measurements. This dissertation introduces novel computational approaches to interrogate how patterns of intercellular interactions are linked with the spatial organization of cells and various cellular contexts. A 3D atlas of cells in Caenorhabditis elegans, a model organism with stereotypically located cells, is used to infer a spatial code underlying cell–cell interactions. To do so, a computational tool called cell2cell was developed, which demonstrated a negative correlation between intercellular distances and inferred interactions. This tool was integrated to a genetic algorithm to identify ligand–receptor pairs informing about the spatial organization of cells. Expanding beyond spatial patterns of intercellular interactions, other cellular contexts shaping interactions can be studied. However, current methods are limited, either disregarding distinct contexts or relying on simple pairwise comparisons. To address this challenge, an unsupervised method using tensor decomposition, called Tensor-cell2cell, is introduced. Tensor-cell2cell identifies context-driven patterns of communication associated with different phenotypic states determined by unique combinations of cell types and ligand–receptor pairs. Thus, Tensor-cell2cell is applied to identify multiple modules associated with distinct communication processes linked to COVID-19 and autism spectrum disorder. By adapting Tensor-cell2cell, different combinations of cellular contexts can be analyzed. For example, autism spectrum disorder is a neurodevelopmental disorder with complex genetic origins that often affects cell–cell interactions in the brain. Thus, single-cell data from brain organoid models of autism is analyzed with an adapted version of Tensor-cell2cell to identify patterns accounting for multiple genetic mutations and developmental stages simultaneously. Key ligand–receptor pairs are identified, with preliminary experiments supporting the patterns detected by Tensor-cell2cell. Finally, Tensor-cell2cell is adapted to include both protein and metabolite ligands as mediators of cell–cell communication. This enables studying temporal dynamics during, for example, brain development and detecting the concerted use of key protein and metabolite ligands by specific interacting cells. Thus, this approach provides insights into complex communication patterns across diverse conditions.