Computational Approaches for Understanding Multi-Area Circuit Mechanisms Underlying Memory-Guided Movements

Memory-guided movement is essential to our everyday activities. Understanding its underlying neural mechanism is critical for developing therapeutic treatments for neurological diseases such as Parkinson’s disease. Memory-guided movements rely on two processes: maintaining short- term memories and executing movements based on these memories. The short-term memory is supported by persistent neural activities coordinated across multiple brain areas, while movement execution relies on the signal sent by medulla, which integrates inputs from both cortical and subcortical areas. We are only at the beginning of understanding how these brain areas interact during these processes through electrophysiology recordings and optogenetics, a full picture of area-to-area coordination remains elusive. A large-scale electrophysiological dataset recently released by Svoboda’s lab (led by my co-mentor) offers an unprecedented opportunity to study the multi-area neural mechanism underlying memory-guided movement. This dataset covers nearly all relevant brain areas across multiple experimental sessions, in which, however, only a subset of brain areas is simultaneously recorded at each session. Therefore, computational methods must be developed to infer the whole network’s behavior from the set of snapshots of subsets of the network provided by multi-session, multi-area Neuropixel datasets. To address this challenge, I propose two complementary aims. In Aim 1 (K99), I will develop machine learning methods to infer simultaneous neural activity patterns from all relevant areas based on recordings from different subsets of areas across sessions. In Aim 2 (K99/R00), I will identify a family of multi-area circuit models with biological constraints and motor outputs to best explain neural and behavioral data across sessions. These circuit models will provide insights into how brain areas coordinate and enable predictions about the most effective optogenetic perturbations to modulate behavior. In summary, this research will not only develop computational framework for analyzing the increasingly ubiquitous large-scale electrophysiological datasets but also provide mechanistic insights into memory-guided movements, with potential applications in treating motor disorders in the long-term. The K99 phase, I will work at the Zuckerman Institute at Columbia university, receiving training in computational neuroscience from Drs. Kenneth D. Miller and Liam Paninski, and training in experimental neurobiology from Dr. Karel Svoboda. Their mentorship and scientific expertise will equip me with skills and knowledge to lead an independent research group in computational neuroscience.