Item

Abstract

During a decision-making task, neurons correlate not only with behavioral task variables such as vision, action selection, and prior expectation, but also overall task responsiveness and various types of body movement. Neural activity correlated with these behavioral task variables and movement may be distributed across brain regions. Here we use non-parametric statistical tests to delineate the spatial distribution of single-neuron activity relevant for these processes across nearly the whole brain. To achieve this, we introduce a dataset of brain-wide Neuropixels recordings from mice performing a visual discrimination task in which they turn a wheel left or right to indicate the location of a presumed visual stimulus. In each recording session, the probability of stimuli appearing on the left vs. right is 80% (or 20%) in blocks of consecutive trials. As of June 2022, the recordings consist of 49292 neurons recorded from 258 brain regions, combining recordings from 12 laboratories.
To quantify the sensitivity of single neurons to the visual stimulus (left vs. right side of visual stimulus), action selection (left vs. right direction of turning), action initiation (pre-movement vs. baseline) and prior expectation (left vs. right block-side), we computed the sensitivity metric of given condition based on the combined-conditions Mann-Whitney U statistic. We then computed the fraction of neurons in each brain region significant for the behavioral task, visual stimulus, action selection and prior expectation, and identified brain regions that are most relevant for these conditions. We found that about 29% of the brain regions are significant for coding of prior expectation. About 30-50% of brain regions are significant for side of visual stimulus/action selection.
To quantify the responsiveness of neural activity in task epoch, we used Wilcoxon rank-sum test to compare firing rates in baseline (pre-stimulus) and different task periods aligned with stimulus onset, first-movement onset or reward. To measure the behavioral movement correlates of single neurons in the entire recording sessions, we applied a time-shift test to compute the significance of Pearson correlation coefficient between binned spiking activity and time-series of behavior variables. We found that in most brain regions more than half of the neurons have significant task responsiveness and correlations to behavioral movement variables.
Overall, we observed widespread coding of movement and task responsiveness but more restricted representations of task-variable coding. These results suggest that neurons across many brain regions are modulated by performance of a behavioral task.