The goal of this project is to develop a mathematical model of the initial processing of olfactory information in the brain, which is performed in the olfactory bulb. General information-theoretic considerations as well as experiments indicate that the main task of the olfactory bulb is to decorrelate odor stimuli. Its ability to do so depends sensitively on a match between the connectivity of the neural network and the odor environment. The project will focus on the ability of the bulbar neural network to adapt its connectivity to changing olfactory environments of the animal, while retaining the ability to discriminate and recognize significant previous odors. To model the plasticity of the network Hebbian long-term potentiation of the synapses will be included as well as the substantial neurogenesis that is an almost unique feature of the olfactory bulb. First the research will aim to establish biophysically plausible synaptic plasticity rules that enable the resulting network to decorrelate its odor inputs effectively. Then networks resulting from neurogenesis and cell death alone will be studied and their ability to decorrelate inputs will be assessed. Finally, synaptic and neurogenetic plasticity will be combined. It is expected that in the full model synaptic plasticity will provide fast fine-tuning of the adaptation within the connectivity framework set by neurogenesis, which in turn will retain significant information about past environments. Mathematically, the model will consist of a co-evolving bipartite network of weighted nodes and links. Computational and asymptotic solution methods will be developed to deal efficiently with the combination of discrete, stochastic evolution of the network structure and continuous deterministic evolution of its weights, involving three different time scales associated with neural activity, synaptic plasticity, and network connectivity, respectively.
A central function of the brain is to extract relevant features from the information it receives from the sensory organs and to make decisions based on new and previously stored information. The brain areas that perform the first information processing play an essential role by providing an internal representation of the environment that facilitates the extraction of relevant information. An important goal in neuroscience is to understand these internal representations and how they depend on the task at hand. In contrast to vision, where great progress has been made in understanding the neural encoding of the visual scenes that an animal is seeing in its natural environment, the processing of odor information in the olfactory system is much less understood. Unlike the visual world the olfactory world is high-dimensional and its characteristic features change significantly over time (seasons, migration, etc.). As the animal adapts to these changes the internal representation of a given odor is likely to change, as well. If this were the case the same rose would be perceived as smelling like one flower in spring and like another flower in fall, say, and the animal would have difficulties to identify odors. The goal of this project is to elucidate how the olfactory system copes with the conflicting tasks of adapting to changing environments and recognizing significant odors reliably in different environments. A central component of the model will be the substantial birth of new neurons (`neurogenesis') that is observed in the olfactory system throughout the life of the animal. The insights gained in this project are likely to be useful for the design of `artificial noses'. The project will also shed light on the role of neurogenesis in the hippocampus, where altered neurogenesis has been associated with mood disorders and neurodegenerative diseases.
