There is significant current interest in understanding how neurons relate to whole-brain networks and how these networks coordinate to give rise to behavior. We have recently shown that we can decompose Local Field Potential voltage signals recorded from depth electrodes implanted in multiple brain regions into ?electrical functional connectomes,? or electomes, which empirically are consistent between individuals in a population and are predictive of behavior. This provides a unique opportunity to use these consistent network properties to anchor and stratify neurons based on their relationship to whole-brain network dynamics, and this approach only uses data that is currently collected or could be collected with minor changes in protocol. This stratification will be consistent across animals, allowing examination of how neurons in different brains relate to whole-brain networks and to behaviors. This will facilitate closer examination of neuron activity by allowing stronger data aggregation across subjects. Neural encoding and decoding models will be required to be faithful at both the individual neuron level and the multi-region voltage dynamic level, constraining neural relationships by known and learned biological properties. This will be accomplished by first building upon our explainable machine learning approach for learning multi-region networks by including historical data to improve the utility of machine learning during pilot studies and creating novel stochastic optimization approaches for real-time mapping of observed data into the networks. Then, we will develop a model capable of embedding neurons into a shared space, where the embedding succinctly reveals how a neuron relates to the electomes. Finally, these vector embeddings will be used to relate neurons to both behavior and the multi-region dynamics in both single-trial and repeated-trial behavioral tasks. Remarkably, a neuron?s embedding and the state of the mesoscale networks can predict the neural encoding a priori to observing behavior and neural firing simultaneously in a new subject.
The overarching goal of this proposal is to learn how neurons? action potentials, long considered to be a fundamental unit of information, relate to whole-brain spatiotemporal voltage patterns and behavior. To uncover this relationship, we will develop novel computational methods capable of learning networks that relate voltage signals from multiple brain regions based upon our previously developed explainable machine learning approach. These networks will then be used to stratify neurons into subtypes consistent across a population of subjects to facilitate the statistical aggregation of data to uncover relationships between multiple scale of neural activity and behavior.
