From Spiking Patterns to Memory Formation: Tools for Analysis and Modeling of Network-wide Cognitive Dynamics of the Brain

The brain routinely integrates polymodal sensory inputs into a coherent representation of events that can be subsequently consolidated into long lasting memories. A long-standing question is how fleeting experiences, encoded by the action potentials of small, sparse neuronal populations, can modify neural networks to produce memories that are both stable and robust to interference over long timescales. This is one example of a general technical hurdle facing neuroscientists. The advent of new recording technologies allows investigators to monitor activity in hundreds or thousands of neurons simultaneously in vivo, during behavior. In theory, this allows neuroscientists to establish links between neural network activity and basic brain functions, e.g., encoding new information during experience. Neuroscientists face two issues in attempting to optimally analyze and characterize these activity patterns. First, neural activity patterns typically are quantified over milliseconds-to-minutes timescales, while behaviors evolve over longer timescales (seconds-to days or even years). Second, it is not obvious what features of network activity and dynamics constitute a “signal” associated with a specific brain function, vs. “noise” which is irrelevant to that function. As capabilities increase for recording larger neuronal populations across longer periods of time, both of these issues become more pronounced. Novel computational tools, which can more rapidly quantify functional connectivity and network dynamics over time, will be critical to address long-standing questions in neuroscience. Network physics, aimed at quantifying mesoscopic and macroscopic network structural and dynamical properties, is ideally suited to this task. Collaborative research between the Zochowski and Aton labs has established a novel framework, built on more rapidly estimating network functional connectivity, to characterize the dynamics of memory encoding and storage. Based on preliminary computational modeling and experimental data analyses, we hypothesize that: 1) neural network dynamics change in a predictable and conserved way immediately following learning, and 2) these dynamic changes are responsible long- lasting memory formation. The aim of this proposal is to develop and test network dynamics metrics which: 1) will apply to our specific neurobiological question, and 2) will be highly useful to neuroscientists addressing similar network-level questions.