Automated characterization of cellular morphology and subcellular expression patterns using spatial transcriptomics and deep neural networks

The development of spatial transcriptomics (ST) techniques, which were named “Nature’s method of the year 2020”, is revolutionizing the analysis of biological tissues. It is able to describe a heterogeneous cellular landscape, while retaining the spatial dimension. High-throughput ST techniques like MERFISH can characterize the cellular transcriptome at subcellular resolutions, however a high demand for computational techniques exists that are able to process and extract meaningful information from the resulting images. Here, we propose the development of novel computational approaches for cell type identification purely based on histological microscopy images, and subcellular expression pattern characterization using MERFISH data and state-of-the-art deep neural networks. This PhD project provides a framework for improved ST data analysis, as well as fundamental insights into cellular morphology and subcellular expression patterns in tissue microarchitectures. The proposed AI-driven cell type identification will leverage MERFISH data for training, but has many potential applications in standard microscopy analysis of histological samples in digital pathology. These methods will be applied in a variety of biological contexts such as: the role of genetic variation in the pathogenesis of Parkinson’s disease and human embryo development, but can potentially be applied to any field of biology.