Computational modeling of retinal development

The development of retinal tissue is based on the well-orchestrated interaction between genetic programs and environmental cues. Genetic instructions produce certain factors, such as extracellular substances, which conversely influence gene expression. Insights into this dynamic, developmental process are of high medical relevance. For instance, a better understanding of the developmental programs of precursor cells would support stem cell-based treatments for the repair and regeneration of the injured or diseased retina. Also, elucidating how neurodevelopmental diseases give rise to severe malformations and impaired function would help devise new clinical approaches to counteract disease progression.

The retina is an ideal candidate for investigating neural development. Because of its peripheral location, it is one of the best-studied parts of the central nervous system (CNS), with much experimental data publicly available (e.g. the synaptic connectivity between individual cells or reconstructions of neuronal morphologies). Additionally, its organisation is simpler as it is composed of only 3 cell layers, in contrast to the 6-layered mammalian neocortex. Furthermore, there is a basic consistency of the retina across species, and so the insights on the retina for specific species are very relevant also for others.

Computer simulations bridge the gap between theory and experiment by providing an alternative way of investigating complex systems, and allow bringing together different lines of research into one coherent framework. However because of limitations in computing performance, models of retinal development have remained on a rather abstract level, and do not incorporate detailed neuronal structure and function.

Here I propose to investigate retinal development based on a novel Neuroinformatics approach, using state-of-the-art software and hardware technology. I want to simulate retinal development by taking into account detailed physical interactions. These comprise mechanical forces between neurons and the secretion and detection of substances in 3D space. I will assess the accuracy and biological plausibility of the simulations with datasets from a number of studies on retinal structure and function. Also, this work will take into account gene expression data obtained from cultures of human stem cells for the growth of transplantable, retinal tissue. By following this detailed simulation approach, I want to derive experimentally testable predictions on the effects of alterations during specific stages of retinal development. These are the main questions that I would like to address in this research:

- How does, starting from a few precursor cells, retinal lamination occur? Here I will incorporate knowledge from gene expression data obtained from cultures of human stem cells. This analysis will allow me to relate these to the interaction with the physical patterning.
- How do neurons develop their structure within the laminated retina? This question requires the investigation of how different developmental cues (e.g. extracellular substances, electrical activity) shape neuronal morphologies.
- How do neurons coordinate their developmental behavior, giving rise to the experimentally measured synaptic connectivity? Again, this problem requires the modeling of dynamic interactions between neurons.

Overall, this project addresses the problem of how bottom-up genetic and top-down developmental processes produce retinal structure. I aim at a detailed computational model of retinal development, offering an in-silico approach for understanding the rise of retinal diseases and putative new treatment routes.

The proposed research aims at a computational characterization of retinal development. I want to simulate the growth of the structural development of the retina, by incorporating results from different studies into a coherent framework. This detailed approach enables a mechanistic explanation for how retinal tissue develops from a few precursor cells, which has high relevance for stem cell-based treatments. This work will use genetic data obtained from retinal differentiation of human embryonic and induced pluripotent stem cells, from the collaboration with Prof. Lako. This is a unique dataset because the samples will be taken at different stages during retinal development. The fact that these are human cells renders these data highly relevant for medical purposes. Based on these data, gene regulatory dynamics will be inferred and related to the structural changes. This model will support the generation of tailored, multilayered neural retinas for drug discovery purposes, disease modeling and transplantation.

Leveraging modern computer hardware and software, this Neuroinformatics project will incorporate very elaborate mechanisms. In contrast to existing approaches to model retinal development, physical forces will be taken into account to simulate cell interactions. The production, diffusion and detection of extracellular substances will enable the modeling of the guidance of neuronal growth processes. Also, the interaction between anatomical and electrophysiological development will be investigated. For example, dendritic growth can depend on electrical activity, as supported by experimental data. Importantly, the simulation is such that every cell behavior is based on information that is available at the location of the process in question. Overall, this detailed and mechanistic model of retinal development will generate experimentally testable hypotheses, and will account for findings such as the specificity of synaptic connections between neurons.

This project is relevant from a number of perspectives, and will have a wide range of beneficiaries (in addition to the academic beneficiaries):

1) Medical applications

Eye health is a major determinant of quality of life. Worldwide, approximately 39 million people suffer from blindness. In the UK alone, the total cost of visual impairment in 2008 has been estimated to be approximately £22 billion (Access Economics, 2009), with an increasing trend. Computational modeling of retinal development has high relevance for the treatment of visual impairment or blindness. Regenerative medicine for the replacement of cells that have been lost, e.g. in retinal blindness, offers promising approaches for treating disorders of the eye. However, the generation of laboratory grown synthetic retina is currently in its infancy. Quantitative models on the relevant developmental mechanisms, and of how and when they shape retinal tissue would help researchers to generate customized retinal tissue in vitro, which could be used in transplantations. Furthermore, elucidating the mechanisms of retinal development will improve our understanding of neurodevelopmental disorders. Such insights into how and when the wrong execution of developmental processes generates associated malformations will support treatments to counteract the disease.

2) Industrial High-Performance Computing (HPC)

This project will benefit from a collaboration with CERN openlab, aiming at a novel software framework for conducting large-scale simulations using hybrid cloud computers. By simulating the computationally very challenging problem of retinal development, my work will demonstrate the power and efficiency of this particular approach. The same software structure can in principle be used for a wide variety of HPC applications, such as for example simulations for industrial engineering.

3) Influence on Society

I am deeply committed to communicating my work to the general public. Since my work will be based on very detailed models in 3D, the results of this project can brought across in an easily understandable way. Visualizations of the simulations are often aesthetically appealing (e.g. my publication in the journal Cerebral Cortex was illustrated on the cover). I will upload results on a personal website and Youtube, in order to bring my research closer to the public. See also "Pathways to Impact" for further information on my plans for public engagement.

4) Animal use in research

Computer simulations provide a third scientific way, in addition to theory and the traditional experimental method. Already now, it has become possible to do experiments in computer simulations that would not be feasible otherwise. Ultimately, with more detailed and accurate models of a system, it will become possible to conduct medical research and development in computer simulations, and so decrease the use of animals in medical research. For example, predictions on highly probable complications or low information gain could be used in order to improve the efficacy of animal experimentation.