Accelerating Biological Discovery with Deep Learning and Spatial Optical Barcodes

Citation

Schwartz, Morgan Sarah (2024) Accelerating Biological Discovery with Deep Learning and Spatial Optical Barcodes. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/55c7-8142. https://resolver.caltech.edu/CaltechTHESIS:05282024-221603734

Abstract

Methodological advances in biology have given us a powerful suite of tools for measuring the state of the cell. Among these methods, next-generation sequencing, including single-cell methods, enables comprehensive measurement of gene expression; however, sequencing-based methods often preclude the collection of other visible phenotypic information. In contrast, light microscopy supports many different measurements that can be acquired in sequential rounds of labeling and imaging because light microscopy does not destroy the sample. Furthermore, light microscopy supports live cell imaging, including the use of fluorescent reporters to observe signaling dynamics in real time. In order to fully understand cellular function, multimodal data collection is needed that encompasses live cell response, end-point phenotypes, and finally perturbations to test the components of relevant signaling networks. In this thesis, I present key advances to create a unified experimental platform for interrogating the cell state. This platform uses light microscopy to collect multimodal measurements of cell state while supporting high-throughput perturbation screening. This platform is supported by a suite of deep learning analysis tools to enable quantitative analysis of these high-dimensional datasets. In Chapter 2, I introduce Caliban, our deep learning method for nuclear segmentation and tracking. In Chapter 3, I present a new method of optical barcodes to enable microscopy-based pooled perturbation screens. Finally, in Chapter 4, I describe preliminary work that leverages the previously described cell tracking and barcoding methodologies to explore the interdependencies of signaling pathway dynamics.

Thesis Files