Computational systems approaches to spatial and multi-omics analysis of tissues and organoid models

Abstract

This thesis focuses on the application and evaluation of computational methods to analyse data generated from the latest spatial and multi-omics technologies to characterise tissues and organoid models.

In the first chapter of this thesis, I evaluated the performance of current state-of-the-art spatially variable gene (SVG) detection methods. This demonstrated that SVGs are biologically informative in understanding spatial gene expression across various tissue contexts, and provided insights into the current methods and recommendations for future method development. Following this, I explored the application of spatial and multi-omics technologies in the characterisation of in vitro 3D neural and retinal organoid development to mimic in vivo human development.

We generated a trans-omic map of early neural organoid development profiling the important kinases which propagate cell identity-specific signalling and downstream gene regulation during neurogenesis. Computational analysis uncovered key signalling cascades and enabled us to modulate kinase signalling toward improving neural organoid generation. Complementary to this work, I characterised retinogenesis under the duress of retinoic acid modulation to perturb retinal cell fate. Using multi-resolution spatial transcriptomics, I investigated the spatiotemporal dynamics of transcriptional activity and its influences on cellular heterogeneity during tissue patterning of retinogenesis. This revealed that retinoic acid signaling affects photoreceptor maturation and retinal interneuron abundance, influencing retinal lamina formation and central retina development, thereby contributing to the increased fidelity of retinal organoids to the human retina.

Together, this thesis showcases the application of computational systems approaches for analysing spatial and multi-omics data in tissues and 3D organoid models.