BMI Special Seminar // Mora Ogando: Causally probing neural codes underlying learning and decision making

Animals must extract relevant features from complex sensory scenes, form stable associations through learning, and retrieve these learned codes to guide decisions. Despite major progress, we still lack a unified understanding of how these processes arise from local circuit interactions. In this talk, I will present a set of causal experiments using two-photon all-optical read and write of neural activity that uncover how specific patterns of excitation and inhibition support sensory and cognitive functions. First, I identify a new mechanism for feature-specific suppression in visual cortex and demonstrate how this circuit improves the accuracy of sensory representations. Second, I show that these feature-selective excitatory inhibitory interactions are remodeled as animals learn to use a visual feature in a task. Finally, I use synthetic activity patterns in behaving mice to control perceptual decisions in real time, revealing which components of the learned neural code are sufficient to drive the expression of the behavior. Together, these results outline a framework for how cortical circuits extract, store, and use behaviorally relevant information.