Imaging and computational platform for integrated analysis of biodata

PROJECT SUMMARY
Here we propose a novel technological platform that enables rapid and unbiased analysis of entire organs at
subcellular resolution. An example of application is to study the developmental glial production (gliogenesis) in
the mouse forebrain. Glial cells (also called glia), the most abundant cells in the central nervous system (CNS),
play key roles in formation and modulation of myelin, synaptic transmission, and more. Many developmental and
adult cases of CNS injury and degeneration such as Alexander disease, multiple sclerosis and Alzheimer’s
disease are also associated with alterations in the development of glia. Unlike neurogenesis, gliogenesis remains
active and responsive throughout adulthood, making gliogenesis a suitable candidate for future regenerative
studies. Therefore, unfolding the mechanisms in which gliogenesis is regulated can provide a way to control the
fate of glial cells and, consequently, the ability to reverse neurodegenerative diseases.
However, the study of gliogenesis during development and adulthood is technically challenging. Since we use
mouse genetic approaches to study gliogenesis mechanistically, our studies necessitate unbiased analyses that
require tissue processing, imaging, collection and storage of large datasets, quantification, and regional
annotations of the entire forebrain across space and time. This is practically impossible to achieve using classical
histology and manual or even semi-automated methods that can take several years per gene. We have
established a pipeline for light sheet fluorescence microscopy (LSFM) of tissue cleared (TC) mouse forebrains.
The pipeline automatically locates and classifies millions of cells by color and morphology in thousands of images
and maps them after registration to established global coordinate system such as the widely utilized Allen Brain
Atlas. Combing further software development and hardware innovation, the level of imaging and unbiased
analytical system is unprecedented and has the potential to be game-changing, allowing full and detailed cellular
and subcellular analyses that have heretofore been unattainable.
Specifically, we will: 1) Optimize a pipeline that implements simultaneous and automated detection and
classification of multiple cell types in TC brains with LSFM imaging. To accelerate the adoption of such a pipeline
in a biology lab, a graphical user interface will streamline the user’s ability to verify the results and to correct for
errors using active learning. 2) Deploy a Deep Design (DD) approach for enhancing optical sectioning and multi-
color acquisition in LSFM. The success of the DD initiative holds the potential to unlock various applications
aimed at enhancing hardware reliant on deep learning for data analysis.
Success of these goals will lead to an automated imaging and analysis platform that is developed to address
questions in an important biological problem.