Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster D in Poster Session D: Wednesday, August 5, 2:00 – 3:45 pm, Kimmel Center, Shorin & Rosenthal Rooms

Inferring the Role of Neuromodulators in Credit Assignment Using Recurrent Neural Networks

Rimjhim Tomar¹, Gichan Lee², Abbas Rizvi², Evan S Schaffer¹; ¹Icahn School of Medicine at Mount Sinai, ²University of Wisconsin-Madison

Presenter: Rimjhim Tomar

How the brain assigns credit to individual neurons during learning remains a fundamental open question. Neuromodulators such as dopamine and noradrenaline are prime candidates for carrying learning signals, yet their diffuse release raises a key question: how do global signals generate selective plasticity at individual cells? One candidate mechanism is that heterogeneous G protein-coupled receptor (GPCR) expression transforms broadcast modulatory signals into distinct per-neuron feedback. However, because recurrent dynamics obscure the relationship between a neuron's activity and the feedback it receives, such receptivity has remained unobservable from population activity alone. Here, we develop a computational framework to infer neuron-specific neuromodulatory receptivity from trial-by-trial changes in recurrent population dynamics, and show that ground-truth receptivity can be accurately recovered even under nonlinearities and imperfect error estimates. Applied to calcium imaging data from motor cortex with paired MERFISH-derived GPCR profiles, this framework opens a path toward identifying which neuromodulators and their cognate receptor combinations determine how individual neurons learn.

Topic Area: Methods, Tools, Theory & Neural Coding