“Using AI to predict mechanisms that regulate branching morphogenesis”
Branching structures are ubiquitous in nature, spanning multiple length scales from river networks and trees to human organs like blood vessels and lung tissues, as well as cellular structures such as neuronal cells and subcellular components like branched actin and microtubule filaments. In neurons, branching patterns are crucial for establishing and maintaining proper connections, which are fundamental to brain functions such as learning, memory, and behavior. Aberrant morphologies often lead to neurological and neurodevelopmental diseases. A large number of molecules have been discovered that play crucial role in neuronal morphogenesis. However, the mechanisms that regulate neuronal branching and the morphological changes induced by molecular players remain poorly understood due to the complexity of the system and the challenges posed by analyzing and comprehending extremely large datasets. To circumvent these issues, employing computational and machine learning tools might provide an alternative powerful research avenue.
The objective of my presentation is to share a combinatorial approach that includes creating a robust generative computational software to accurately simulate the development of neuronal branching morphologies and a deep learning tool trained on realistic synthetic images generated by these simulations to predict the underlying mechanisms leading to specific morphologies, with a particular focus on molecular cues. The proposed work will not only deepen our understanding of neuronal morphogenesis but also provide tools for predicting and manipulating neuronal morphology to treat neurodevelopmental and neurodegenerative diseases. Furthermore, this model can be easily generalized into other branching systems.
Hosts: Naomi Gluck & Nathan Suri