CRCNS US-German-Japanese Research Proposal: Ensemble modeling of a visual system connectome

About This Grant

What you see, think, and do is determined by the electrical activity of cells in your brain. These cells are called neurons. Some neurons have electrical activity that is externally determined by the environment, such as neurons in the eye that sense light. However, most neuronal activity is internally generated by neuron-to-neuron signals sent and received at anatomical communication sites called synapses. Understanding how the brain works therefore amounts to discovering how neuronal activity is generated by the brain’s immense number of synaptic connections. For decades neuroscientists have tried to do this using mathematical models called neural networks, but a lack of experimental data has left it unclear whether these models are accurate enough. This Multilateral Research Project combines mathematical modeling, neuronal activity measurements, and synapse-resolution neuroanatomy to build and test biologically realistic neural network models of visual functions. The project capitalizes on recent experimental breakthroughs to build testable models and benchmark a general theoretical framework for modeling the brain and making experimental predictions. It will produce advanced theoretical methods, better annotated datasets, and new experiments that will be freely shared with the scientific community and published in publicly accessible forms. The project heralds a new era of neural networks research, and the research team will help other scientists enter this exciting new field by organizing workshops at mainstream conferences. The work spans three continents (USA, Germany, Japan) and provides rich educational opportunities for undergraduates, graduate students, and postdoctoral fellows to learn the value of multidisciplinary collaboration and international partnership. Quantitatively linking synaptic connectivity and neuronal activity is fundamental towards understanding the brain as a neural network. The three Principal Investigators all provide complementary and indispensable expertise towards this goal. The theoretical foundation for the proposal is a mathematical framework developed by the US Lab for comprehensively analyzing all synaptic weight matrices that allow a neural network model class to perform a specified computation, an approach called ensemble modeling. Ensemble modeling allows one to pinpoint strong predictions that hold consistently across the ensemble, and testing these predictions can validate or rule out an entire model class. These tests are possible in the zebrafish optic pretectum because the German Lab has recently acquired and published a Function-Linked (FuL) connectomics dataset that combines cellular-resolution functional imaging and synapse-resolution connectomics in the same brain. New analyses of this FuL connectomics dataset will allow tests of many theoretical predictions, but critical information on whether synaptic connections are excitatory or inhibitory is missing. The team will therefore use the Japanese Lab’s Function-Guided Inducible Morphological Analysis (FuGIMA) tool to characterize the functional response properties and morphologies of excitatory and inhibitory neurons in the pretectum. They will use these results to estimate synapse signs by the neurotransmitter type of each neuron in the FuL connectomics dataset. This proposal uses tightly integrated feedback loops between theory and experiment to address central neuroscience problems that are otherwise inaccessible. It illustrates a paradigm for extracting theoretical understanding from big data and will encourage theorists to embrace synapse-resolution systems neuroscience. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.