A Computational Framework for Investigating mRNA Localization Patterns in Pancreatic Beta-Cells During Type 1 Diabetes Progression

Abstract

Spatial transcriptomics improves transcriptomic studies by incorporating RNA localization information, which provides a more profound insight into cellular functions, interactions between cells, and their reactions to external stimuli. Single-molecule fluorescent in situ hybridization (smFISH) is a commonly utilized technique in spatial transcriptomics that allows for the accurate visualization of mRNA distribution in cells. This method aids in the quantitative evaluation of mRNA localization patterns by utilizing various physical properties, thereby illuminating processes such as transcription, nuclear export, and localized translation. Nevertheless, existing computational approaches for analyzing smFISH images often have constraints, concentrating primarily on cellular expression or specific biological contexts while overlooking broader physical analysis. In my PhD research, I created STProfiler, a comprehensive tool aimed at an unbiased physical examination of mRNA distribution. STProfiler includes an image analysis workflow that processes raw biological images to effectively detect mRNA and nuclei. It also employs machine learning techniques to biologically interpret mRNA spatial characteristics and categorize cells based on these features. My dissertation illustrates the use of STProfiler in multiple studies investigating the transcriptomic profiles of β-cells during the progression of type 1 diabetes (T1D), uncovering spatial transcriptomic diversity in β-cells. These investigations involve analyzing mRNA clusters and stress granules in pancreatic β-cells, measuring the physical characteristics of mRNAs linked to cellular stress and inflammation in mice developing T1D, evaluating the rise in HLA-DMB mRNA spliced variant in T1D, and exploring miRNA as a potential biomarker for T1D. Furthermore, STProfiler has also proven beneficial in tissue-wide spatial transcriptomics by creating masks for nuclei and cells from biological images and assigning mRNA transcripts to develop subcellular expression profiles. This capability allows for more thorough bioinformatic evaluations. In summary, STProfiler serves as a robust tool for both cell- and tissue-level spatial transcriptomics, offering an unbiased platform for researchers to investigate complex transcriptomic variations within cells.