Machine learning tools to evaluate hiPSC organoid modeling of human brain development

About This Grant

Project Summary Human pluripotent stem cells (hPSCs) have emerged as a powerful tool for generating 3D organoids, enabling the study of human development and disease. These organoids can closely mimic in-vivo cellular context, making them valuable for investigating biological mechanisms. However, the extent to which gene regulation and other cellular and molecular mechanisms are preserved between in-vivo and in-vitro systems, especially in specific cell types, remains uncertain. Recent advancements in single-cell technologies offer opportunities to answer this question, there is a need for effective computational tools for comparing organoids and brain data. Our recent machine learning tool, Brain and Organoid Manifold Alignment (BOMA) successfully integrated eight published single-cell RNAseq (scRNAseq) datasets to uncover shared (or specific) developmental trajectories between human organoids and brains. However, BOMA is limited to scRNAseq. Simultaneous profiling of gene expression and chromatin accessibility of the same cell (single-nucleus multiomics or snMultiomics) allows linking putative regulatory elements to genes thus providing deeper insights of cell-type gene regulatory mechanisms in the developing brain compared to scRNA-seq alone. In fact, NIH funded consortia such as BRAIN Initiative is generating a large amount of snMultiomic data of human brains from prenatal development to adults. Although single cell multiomic data of organoids is currently scarce, we anticipate that more snMultiomics data will be generated for organoids in near future. This project aims to develop machine learning methods and tools for comparative snMultiomics analysis between organoids and brains, enabling the interpretation of developmental gene regulatory mechanisms at cellular level across in-vivo and in-vitro systems. Aim 1 will develop a machine learning method, Brain Organoid Manifold Alignment by Multiomics data (BOMAM), to uncover conserved and divergent developmental stages in brain and organoids. Aim 2 will perform gene regulatory network prediction and analysis to evaluate the fidelity of current organoid protocols. Aim 3 will develop open-source tools for comprehensive evaluation of brain organoids, designed for general-purpose use. In sum, our tools will enhance the efficiency and integration of brain-organoid analyses, especially for biologists and neuroscientists, leading to a deeper understanding of brain cells and their functional characteristics in development.