Novel computational methods for characterizing cell state and cell-cell interactions using spatial transcriptomics with an AI-assisted experimental design

Spatial transcriptomics technologies are improving our understanding of how the spatial organization of cells influences their cellular identity. Microenvironmental and direct cell-cell interactions play a crucial role in tumour biology, neuroscience, macrophage-driven homeostasis and many other biological processes. Most targeted spatial transcriptomics technologies require carefully selected gene panels which will determine the successful outcome of such experiments. Here we propose a novel data-driven approach for experimental design of ISH-based spatial transcriptomics assays to maximize their informative potential. To this end we will utilize deep neural network architectures to improve feature selection. We will leverage such experimental designs to develop novel downstream analysis pipelines for in-depth characterization of cellular microenvironments and cell-cell interactions. This, together with an automated pipeline for high-throughput probe design, will allow researchers to reduce labour/costs and improve experimental output and biological insights gained from spatial experiments. These methodologies will be applied to multiple collaborations and biological contexts. First in the context of macrophage biology, to investigate the role of microenvironment in immune memory. Secondly, the role of breast tumour microenvironment and cell-cell communication in the therapy-response of immune checkpoint blockade treatments will be investigated.