UC San Diego

Cells encode copies of genomic information in RNA molecules and traffic them to distinct subcellular destinations. These RNA molecules serve a milieu of functions including encoding protein sequences for local translation, regulating cellular functions, or aiding in intercellular communication. The breadth of subcellular RNA functionality is not well understood, and our capability of scaling our knowledge of spatial RNA biology to the entire transcriptome has, until now, been an inconceivable undertaking.In this dissertation, I describe our work developing a suite of software tools and novel spatial genomic methodologies, creating a unified vertically integrated stack of next-generation spatial transcriptomic technology. We developed RNAforest, RNAflux, and RNAcoloc, three first-in-class and best-in-class software algorithms for the statistical spatial analysis of RNA subcellular localization, wrapping them into a toolkit called Bento. Each of these tools enable different approaches to RNA distribution analysis, including supervised and unsupervised subcellular RNA patterning recognition, and compartment specific gene-gene pairwise colocalization. We demonstrated the utility of Bento by discovering organelle-specific RNA enrichment modulation by the chemotherapeutic Doxorubicin in Cardiomyocytes. We further utilized our philosophy of using RNA distribution to compute on cellular space by training a generative pix2pix-HD GAN model and a CellPose cyto2torch segmentation model to create a best-in-class cell segmentation model, RNA-masala, that computes cell boundaries in 3D accurately from RNA coordinates alone. RNA-masala is highly performant, leveraging GPU tensor computation making it feasible to segment single cells in 3D from large tissue spatial transcriptomic datasets such as a gigapixel whole mouse pup Xenium dataset. Finally, we developed SPACE-seq, a next-generation spatial transcriptomic method that achieves a Moore’s Law scale inflection point in plexity (untargeted whole-transcriptome, isoform resolution), cost, as well as data dimensionality combining single-molecule imaging with single-molecule sequencing. SPACE-seq allows the measurement of the location and sequence of every unique molecule in every single cell in a tissue section, unlocking multi-scale RNA biology. Altogether, Bento, RNA-masala, and SPACE-seq take spatial transcriptomics takes RNA biology to new heights and introduces the possibility for creating digitized models of cell biology.