Computational approaches for spatialomics data: construction, integration and analysis of biomolecular atlases

Cellular function and behavior are governed not only by molecular composition but also by a cell's native environmental context. By leveraging in-depth molecular quantification and high-resolution imaging methods to study microenvironmental composition and arrangement, we can continue to characterize biological systems and decipher the complex interactions that collectively provide living systems with their function. The development of computational methods that support experimental efforts and analysis in such studies is an ongoing area of research. In this thesis, I investigate how computational frameworks can yield deeper insights into biology by focusing on constructing and integrating biomolecular spatialomics data. To contextualize the work, I begin with a discussion of the relevance of spatial biology, followed by a brief review of the challenges and applications of modern techniques that quantify molecules in tissue.

In Chapter I, I introduce a novel framework, Tomographer, that is designed to enable spatialomics using any bulk measurement technique, thereby overcoming challenges associated with specialized imaging-based approaches. By integrating concepts from compressed sensing and probabilistic modeling, Tomographer leverages quantitative molecular measurements from tissue sections sliced at multiple angles, generating accurate 2D maps of molecular distributions. The results demonstrate Tomographer's versatile applications and its potential to broaden molecular anatomy studies in contexts where tissue availability or imaging capabilities are constrained.

In Chapter II, I address challenges of high-resolution mass spectrometry imaging (MSI) pipelines by developing a method, the Unified Mass Imaging Analyzer (uMAIA), to process and integrate large MSI datasets. The method represents a significant advance in spatial metabolomics, enabling the creation of consistent metabolic atlases and revealing previously unrecognized patterns in lipid distributions across tissues and entire organisms. I apply uMAIA to trace lipid distributions across four developmental stages of the embryonic zebrafish, producing the first 4D lipid atlas and revealing fine-scale "lipid territories" that are spatially and developmentally regulated.

In Chapter III, I generate the first 3D lipid reference atlas for the adult mouse brain. While the brain has primarily been described in terms of its anatomy, circuitry and transcriptomic profile, lipid distributions have been more difficult to quantify and visualize. In this work, I leverage advancements in MSI techniques and data processing to map 410 lipid specie distributions in over 100 coronal sections of the mouse brain. Analysis suggests that lipids provide a refined molecular architecture of the brain, emphasizing the potential of this class of molecular species to advance our understanding of brain function and pathology.

In this thesis, I produce rich descriptions of tissue heterogeneity and answer different b