Computational modelling of neural layer formation
by Roman Bauer
16:00 (40 min) in USB G.003
One of the most characteristic properties of mammalian neural systems is the presence of a layered structure comprising different cell types. This architecture exists in cortex, in the retina, in the hippocampus and many other parts of the central nervous system (CNS). The developmental origin of this common architecture has been the research focus of substantial experimental efforts. However, only few studies provide a quantitative or computational framework for neural layer formation.
In this talk, I will present recent work on agent-based computational modeling of neural layer formation. I will first show how, starting from a small pool of undifferentiated precursor cells, the development of a cortical, layered structure in 3D physical space is simulated. This agent-based model is a composition of a small number of fundamental mechanisms of biology, such as cell proliferation, differentiation, migration and diffusion of chemicals in extracellular space. The temporal coordination of these behaviors is specified through a gene regulatory network model that is instantiated within each cell. The state of the cells is dynamically determined by the interaction of intracellular as well as extracellular processes.
Our results show that apoptosis strongly improves the layer architecture, and enables a wide range of layer thicknesses as measured in human and mouse cortex. Moreover, the characteristic inside-out development of cortical layer formation is recapitulated. Importantly, the simulation of apoptosis is accomplished via two distinct apoptotic processes that require different kinds of input. The overall model complexity is reduced due to the presence of repetitions of similar behaviors. Moreover, characteristic features of certain neurodevelopmental disorders can be accounted for by changes to the gene-type rules of our model, which gives rise to hypotheses on the origins of these disorders.
Finally, I will present some first results towards the simulation of retinal layer formation, including the analysis of immunostaining and RNAseq datasets from the human developing retina.