Federal Project Grant Award Summary

The University of Michigan received a $325,904 Project Grant from the National Institute on Drug Abuse (NIDA) under the Drug Use and Addiction Research Programs (CFDA 93.279), effective March 1, 2026, through December 31, 2028. This award supports the development of innovative machine learning methods for real-time modeling and perturbation of neural manifolds shaped by external stimuli. The project delivers three core research products: (1) real-time dimensionality reduction algorithms that construct low-dimensional neural manifolds incorporating external stimulus information; (2) probabilistic models that predict the effects of neural stimulations on latent neural dynamics; and (3) integrated optimization methods for jointly shaping ongoing neural dynamics through coordinated external stimuli and direct neural stimulation patterns in real time.

These deliverables will be validated through in vivo experiments in larval zebrafish employing high-dimensional visual stimuli and holographic optogenetic photostimulation techniques. The resulting generalizable machine learning methods are designed to automatically learn stimulus-shaped neural dynamics, enabling researchers to causally test competing hypotheses of neural circuit function and advance understanding of how external stimuli drive neural activity. This research addresses fundamental gaps in systems neuroscience by providing tools to dynamically manipulate neural activity in ways conditioned on individual brain states, tasks, and environmental contexts.