【Seminar】"A closed-loop framework for investigating sensorimotor predictions and errors in head-fixed mice" by Dr. Florin Albeanu

Date

Location

Seminar Room C209, Center Building

Description

Talk title: "A closed-loop framework for investigating sensorimotor predictions and errors in head-fixed mice "

We all know the startling feeling of sitting on a ‘seat’ that is just an inch lower than usual! This brief mismatch between reality and expectations reveals a fundamental process that occupies our brain: predicting the consequences of our actions and verifying whether current observations match expectations. Growing evidence suggests that this continuous tally of sensory inputs and predictions is at the core of active perception and represents a canonical cortical computation. However, mechanistically, little is understood how sensory inputs get evaluated against internal predictions.

To quantitatively study sensorimotor predictions both at the behavioral and circuit-level, we developed a novel closed-loop task, Smellocator, wherein head-fixed mice steer the real-time, left-right location of an odor source by controlling a light-weight lever with their forepaws. Briefly, in this task, the odor source is initialized at different distances in each trial. Mice learn to move the lever such as to center the odor in front of their nose to obtain water rewards. Mice learn this task fast (~2 weeks) and perform many trials (~600-800 per session) with exquisite motor precision.

In this manner, we link precise motor actions (lever movement) to well-defined sensory expectations (odor location). We subsequently violate these learnt expectations via online sensory feedback perturbations and interrogate both the behavioral and neuronal signatures of sensorimotor predictions. In one such perturbation, we abruptly displace the odor mid-trial, away from the snout. Strikingly, without explicit training, mice readily counter this perturbation and re-center the odor via precise and predictable movements. The amplitude and direction of these corrective movements reveal the animal’s exact beliefs about the lever-odor function. By employing a diverse repertoire of both brief and long-term perturbations, we can systematically reverse engineer the animal’s beliefs at various time-scales within and across trials.

In parallel, we assessed whether cortical representations are shaped by movement-related expectations by comparing responses to the same stimuli in closed-loop versus passive conditions. In the olfactory cortex, we find nearly equal proportions of expectation-invariant (sensory) and expectation-modulated (sensorimotor) neurons. While a simple linear-non-linear model describes well the responses of ‘sensory’ neurons as a function of instantaneous odor state (identity, location) and active sampling (sniffing), the responses of ‘sensorimotor’ neurons are poorly accounted for by the feedforward drive alone. Integrating in our model estimates of sensorimotor mismatch, as inferred from the animal’s behavior, enables us to identify neurons that compare feedforward sensory input with movement-related predictions.

Overall, here we present a framework that combines novel closed-loop perturbation-based behaviors with neuronal analysis to investigate the circuit elements that compute sensorimotor predictions and errors.

Biography:
Florin Albeanu is a Professor at Cold Spring Harbor Laboratory (CSHL). His laboratory focuses on understanding sensorimotor transformations and on investigating the neural substrates of olfactory perception. Current research investigates the nature of odor space, the flow of information within the olfactory stream, and the circuit mechanisms of internal models (sensorimotor predictions and error signals) in closed-loop olfactory and visual environments. Florin was undergraduate at MIT, did a Ph.D. in Neuroscience with Venki Murthy and Markus Meister at Harvard University, and started his lab as a CSHL Fellow. He served as instructor for the CSHL course on Imaging Structure and Function in the Nervous System and is one of the organizers of the Transylvanian Experimental Neuroscience Summer School (TENSS).

All-OIST Category: 

Subscribe to the OIST Calendar: Right-click to download, then open in your calendar application.