Neural computations underlying flexible control of behavioral strategies and problem-solving

/ABSTRACT Animals exhibit a remarkable array of flexible behaviors. Birds alternate between caching and retrieving food based on availability; rats reroute when familiar paths are blocked; humans revise strategies mid-game in chess. This ability to flexibly switch strategies or generate new solutions is central to intelligent behavior and is often impaired in neuropsychiatric disorders such as autism spectrum disorder and schizophrenia. Prior research has yielded key insights into what supports such cognitive flexibility: internal models of the world, including spatial and episodic knowledge encoded in the hippocampus (HPC) and abstract rules encoded in the prefrontal cortex (PFC). However, we still lack a mechanistic understanding of how the brain engages these models in real time to guide strategy switching and problem-solving. This proposal addresses this gap by identifying internal strategy states—latent variables computed by the brain that track the currently active policy for selecting goal-directed actions—and by dissecting the neural computations that encode, update, and drive transitions between these states. I will combine large-scale electrophysiology with closed-loop optogenetics in freely behaving rats performing strategy-switching and problem-solving tasks. I will assess behavioral and neural data by integrating two complementary theoretical frameworks: (i) reinforcement learning and Bayesian inference to formalize latent behavioral strategies and valuation processes; and (ii) dynamical systems modeling to uncover how neural population activity implements these cognitive operations. I will test the central hypothesis that the flexible control and generation of strategies arise from structured population dynamics in medial PFC (mPFC) implementing computations that: (i) direct HPC to simulate future scenarios that inform strategy switching (Aim 1); (ii) track and update strategy values to determine when to switch (Aim 2); and (iii) integrate input from the orbitofrontal cortex to select among multiple strategies and generate new solutions (Aim 3). By causally linking neural dynamics to strategy switching, selection, and generation, this work will reveal algorithmic and implementational principles of cognitive flexibility, laying the groundwork for my long-term goal: to elucidate the division of computational labor across PFC subregions and their interactions with subcortical regions (e.g., thalamus) during multi-strategy problem-solving. The K99 phase will support my transition to independence through training in multi-region, high-density electrophysiology coupled with real-time optogenetics, as well as advanced behavioral and dynamical systems modeling. I have assembled a mentorship team (Drs. Loren Frank, Joshua Berke, and Nathaniel Daw) and collaborators (Drs. Vikaas Sohal and Scott Linderman) with complementary expertise spanning experimental, technological, and theoretical domains of systems neuroscience. This award will also provide professional development in lab management, leadership, scientific communication, and grant writing, which will position me to launch an independent research program focused on the neural basis of intelligence and creative behavior. Project Number: 1K99MH144375-01 | Fiscal Year: 2026 | NIH Institute/Center: National Institute of Mental Health (NIMH) | Principal Investigator: Xulu Sun | Institution: UNIVERSITY OF CALIFORNIA, SAN FRANCISCO, SAN FRANCISCO, CA | Award Amount: $117,747 | Activity Code: K99 | Study Section: Special Emphasis Panel[ZRG1 ICN-L (92)] View on NIH RePORTER: https://reporter.nih.gov/project-details/11353013