Cellular interactions underpin all biological processes and offer unprecedented insight into mechanisms of action in steady state and disease. The advent of single-cell and spatial technologies has allowed us to resolve these interactions across time and space unveiling novel pathways in infectious and inflammatory disease, yet interpretation and visualisation remains challenging in these multi-faceted high-dimensional datasets. This thesis develops and applies computational and visual approaches to infer, prioritise, and validate cell–cell communication (CCI) in such contexts, demonstrating leveraging spatial information allows us to hone in on biological hypotheses and reduces false positives in ligand–receptor analyses. In Chapter 1, I analyse lethal COVID-19 in a Malawian cohort using histology, high-dimensional imaging, and single-cell transcriptomics from lung, blood, and nasal tissues, integrated with datasets from Northern Hemisphere cohorts. This cellular interaction analysis reveals distinct immune drivers in our cohort: an interferongamma programme in lung-resident alveolar macrophages in Malawi contrasted with type I/III interferon responses in blood-derived monocytes reported in USA/European cohorts. These results provide mechanistic insight into fatal disease in an under-represented population, and highlight the value of context-aware cellular inference and validation. In Chapter 2, I introduce cellXplore, a Flask–React interactive visualisation web tool that unifies widely used CCI packages and leverages single cell RNA sequencing with spatial transcriptomics to investigate computed cellular interactions. Through interactive, point-and-click workflows, cellXplore streamlines analysis, allowing customisable interactive plots, and prioritises spatially plausible interactions by overlaying ligand–receptor expression with co-localisation of spatial gene expression. I present three end-to-end user workflows using single cell and spatial transcriptomics data from a 10X Visium parasitic infection and a 10X Xenium breast cancer dataset to show indirect spatial validation of cellular interactions can be utilised in a user-friendly manner. Lastly in Chapter 3, I extend cellular communication inference to complex datasets, validating cellular interactions and key drivers of inflammation leveraging immunohistochemistry and spatial transcriptomics. A multifactor macrophage–fibroblast atlas spanning four tissues and inflammatory states reveals conserved tissue-resident myeloid–stromal circuits through a APOE+/SPARC+ - SPP1+ axis that underpins inflammation alongside tissue-specific crosstalk reflecting organ microenvironments. A second study applies 10X Visium to intestine ’gut-rolls’ across four time points of Heligmosomoides polygyrus infection. The analysis uncovers epithelial and immune programs associated with granuloma formation, stem-cell reprogramming, and parasite driven immunomodulation within a distorted tissue landscape, with cellular interactions validated in the spatial context. Together, these studies shine a spotlight on the power of spatially-aware cellular interaction inference providing insight into COVID-19, tissue-resident myeloid–stromal communication during inflammation, and helminth infection in addition to a novel visualisation tool to unlock new insights from cellular interaction data.
Olympia Melek Hardy (Thu,) studied this question.